Download This PDF is a selection from a published volume from... Bureau of Economic Research Volume Title: Housing and the Financial Crisis

This PDF is a selection from a published volume from the National
Bureau of Economic Research
Volume Title: Housing and the Financial Crisis
Volume Author/Editor: Edward L. Glaeser and Todd Sinai, editors
Volume Publisher: University of Chicago Press
Volume ISBN: 978-0-226-03058-6
Volume URL: http://www.nber.org/books/glae11-1
Conference Date: November 17-18, 2011
Publication Date: August 2013
Chapter Title: International Capital Flows and House Prices:
Theory and Evidence
Chapter Author(s): Jack Favilukis, David Kohn, Sydney C.
Ludvigson, Stijn Van Nieuwerburgh
Chapter URL: http://www.nber.org/chapters/c12626
Chapter pages in book: (p. 235 - 299)
6
International Capital Flows
and House Prices
Theory and Evidence
Jack Favilukis, David Kohn, Sydney C. Ludvigson,
and Stijn Van Nieuwerburgh
6.1
Introduction
The last fifteen years have been marked by a dramatic boom- bust cycle in
real estate prices, a pattern unprecedented both in amplitude and in scope
that affected many countries around the globe and most regions within the
United States (figure 6.1). Over the same period, there were economically
large fluctuations in international capital flows. Countries that exhibited
the largest house price increases also often exhibited large and increasing
net inflows of foreign capital that bankrolled sharply higher trade deficits.
Economists have debated the role of international capital flows in explaining
these movements in house prices and asset market volatility more generally.
A common hypothesis is that house price increases are positively related to
a rise in the country’s net foreign inflows, either because they directly cause
house price increases (perhaps by lowering real interest rates), or because
other factors simultaneously drive up both house prices and capital inflows.
Jack Favilukis is a lecturer of finance at the London School of Economics. David Kohn is a
PhD student in economics at New York University. Sydney C. Ludvigson is professor of economics at New York University and a research associate of the National Bureau of Economic
Research. Stijn Van Nieuwerburgh is professor of finance and director of the Center for Real
Estate Finance Research at Stern School of Business at New York University and a research
associate of the National Bureau of Economic Research.
Prepared for the National Bureau of Economic Research “Housing and the Financial Crisis”
conference, November 17– 18, 2011, Cambridge, Massachusetts. We are grateful to seminar
participants at the NBER “Housing and the Financial Crisis” conference November 2011 and
the HULM conference April 2012, and to John Driscoll, Victoria Ivashina, Steven Laufer,
and Nikolai Roussanov for helpful comments, and to Atif Mian and Amir Sufi for data. For
acknowledgments, sources of research support, and disclosure of the authors’ material financial relationships, if any, please see http://www.nber.org/chapters/c12626.ack.
235
236
Fig. 6.1
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
Price-rent ratio in the United States
Notes: The figure plots an aggregate price- rent ratio index for the United States from 1975:Q4
to 2010:Q4. Rent is rent for primary residence, constructed from the Shelter component of
the Consumer Price Index for all urban consumers, SA, last month of each quarter. Data
available from the Bureau of Economic Analysis of the US Department of Commerce. Price
is the Core Logic National House Price Index (SA, Jan. 2000 = 100). The price- rent ratio has
been normalized to equal the level, in 1975:Q4, of the quarterly price- rent ratio constructed
from the flow of funds housing wealth and National Income and Products data on housing
consumption.
In this chapter, we study both theory and evidence that bears on this hypothesis, focusing on the unprecedented boom- bust cycle in housing markets that took place over the last fifteen years.
We argue that changes in international capital flows played, at most, a
small role driving house price movements in this episode and that, instead,
the key causal factor was a financial market liberalization and its subsequent reversal that took place in many countries largely independent of
international capital flows. Financial market liberalization (FML hereafter)
refers to a set of regulatory and market changes and subsequent decisions
by financial intermediaries that made it easier and less costly for households to obtain mortgages, borrow against home equity, and adjust their
consumption.
By contrast, we argue that net capital flows into the United States over
both the boom and the bust period in housing have followed a largely independent path, driven to great extent by foreign governments’ regulatory,
reserve currency, and economic policy motives. Consider the value of foreign
holdings of US assets minus US holdings of foreign assets, referred to hereafter as net foreign asset holdings in the United States, or alternatively, as
the US net liability position. A positive change in net foreign asset holdings
International Capital Flows and House Prices
237
indicates a capital inflow, or more borrowing from abroad.1 As we show in
the following, from 1994 to 2010, only the change in net foreign holdings
of US securities (equities, corporate, US Agency and Treasury bills and
bonds) shows any discernible upward trend. Moreover, among securities, the
upward trend has been driven entirely by an increase in net foreign holdings
of US assets considered to be safe stores- of-value, specifically US Treasury
and Agency debt (referred to hereafter simply as US “safe” assets). Yet
inflows into these securities, rather than declining during the housing bust,
have on average continued to increase. Importantly, foreign demand for US
safe assets is dominated by Foreign Official Institutions, namely government entities that have specific regulatory and reserve currency motives for
holding US Treasuries and other US- backed assets, and that face both legal
and political restrictions on the type of assets that can be held (Kohn 2002).
Such entities take extremely inelastic positions, implying that when these
holders receive funds to invest, they buy US Treasuries regardless of the
price (Alfaro, Kalemli-Ozcan, and Volosovych 2011; Krishnamurthy and
Vissing-Jorgensen 2007).
We interpret these recent events through the lens of the theoretical models in Favilukis, Ludvigson, and Van Nieuwerburgh (2010), henceforth
FLVNa, and Favilukis, Ludvigson, and Van Nieuwerburgh (2011), henceforth FLVNb. These papers study the economic consequences of both the
US FML (and its reversal) and, at the same time, empirically calibrated
fluctuations in net capital inflows into the US riskless bond market. The
model environment is a two- sector general equilibrium framework with
housing and nonhousing production where heterogeneous households face
uninsurable idiosyncratic and aggregate risks. Given the assets available in
the model economy and collateralized financing restrictions, individuals
can only imperfectly insure against both types of risk. We argue that these
frameworks can account for the observed boom- bust pattern in house prices
simultaneously with the continuing trend toward greater net capital inflows
into US securities over both the boom and the bust.2 Fluctuations in the
model’s price- rent ratio are driven by changing risk premia, which vary
endogenously in response to cyclical shocks, the FML and its subsequent
1. What we have defined as net foreign asset holdings, or the US net liability position, is equal
to the negative of the US net international investment position in the US Bureau of Economic
Analysis balance of payments system. A country’s resource constraint limits its expenditures on
(government and private) consumption and investment goods, fees, and services, to its domestic
output plus the change in the market value of its net liabilities (minus the change in the net international investment position). Thus a country’s ability to spend in excess of domestic income
in a given period depends positively on the change in its net foreign liabilities.
2. There was considerable volatility in the changes of net foreign asset holdings in the United
States during the financial crisis in 2008 and 2009. Nevertheless, we show in the following that
the changes in holdings were still higher at the end of the sample in 2010 than they were at the
peak of the housing boom in 2006.
238
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
reversal, and capital inflows. In FLVNa, house prices rise in the boom period
because of a relaxation of credit constraints and decline in housing- related
transactions costs, both of which reduce risk premia. Conversely, the reversal of the FML raises housing risk premia and causes the housing bust.
In contrast to the FML, an inflow of foreign money into domestic bond
markets plays a small role in driving home prices in the models of FLVNa
and FLVNb, despite its large depressing influence on interest rates. The
reason is that a capital inflow into the safe bond market—by itself—raises
risk premia on housing and equity, as domestic savers are forced out of the
safe bond market and into risky securities. (We emphasize the words “by
itself ” here because this increase in risk premia is more than offset by the
simultaneous decline in risk premia during the boom caused by the FML,
as discussed later.) At the same time, the capital inflow stimulates residential
investment and an expected increase in the housing stock. So while low
interest rates in isolation tend to raise home prices, these general equilibrium
consequences tend to reduce them, thereby limiting the scope for a capital
inflow to increase home prices significantly. It follows that the sharp rise in
price- rent ratios during the boom period must be attributed to an overall
decline in risk premia and not to a fall in interest rates. Many alternative
theories that can account for the positive correlation between house prices
and capital inflows in the boom period are not able to explain the bust
period, in which house prices collapsed but inflows into countries like the
United States continued.
By FML we mean an outward shift in the broad availability of credit, at
any given initial level of credit demand and borrower quality. This includes,
as in the US housing boom, an increase in maximal loan- to-value (LTV)
ratios (e.g., the fraction of loans with combined—first and second—mortgage LTV ratios above 80 percent or 90 percent, an increase in the prevalence
of new mortgage contracts (option- adjustable- rate mortgages [ARMs],
interest- only and negative amortization loans, loans to households with
low FICO scores), a reduction in documentation requirements (asset and
income verification), a rapid increase in the use of private label securitization, and a reduction in fees (as well as in time and effort) associated
with refinancing a mortgage or obtaining a home equity line of credit. The
widespread relaxation of credit standards is well- documented (see the following discussion). Consistent with this evidence, microeconomic evidence
in Mian and Sufi (2009) shows that mortgage credit expansion and house
price growth in the boom were concentrated in areas with a large fraction of
subprime mortgages and securitization of these mortgages, and not in areas
with improved or improving economic prospects. Thus, this component of
credit availability to households—accompanied by government deregulation of financial institutions and widespread changes in the way housing
assets were financed and traded—appears to have fluctuated, to great extent,
independently of current and future economic conditions.
International Capital Flows and House Prices
239
But credit availability can also change endogenously in response to fluctuations in the aggregate economy and to revisions in expectations about
future economic conditions, including house price growth. This information
is reflected immediately in collateral values that constrain borrowing capacity. As in classic financial accelerator models (e.g., Bernanke and Gertler
1989; Kiyotaki and Moore 1997), endogenous shifts in borrowing capacity
imply that economic shocks have a much larger effect on asset prices than
they would in frictionless environments without collateralized financing
restrictions. Both exogenous and endogenous components of time- varying
credit availability to households are operative in the model of FLVNa.
While endogenous fluctuations in credit availability are clearly important
in theory, it is unclear how quantitatively important they have been empirically, especially in the recent housing boom- bust episode. Some researchers have argued that credit availability is primarily driven by the political
economy, and in particular by political constituencies that influence bank
regulation related to credit availability (e.g., Mian, Sufi, and Trebbi 2010;
Rice and Strahan 2010; Boz and Mendoza 2010; Rajan and Ramcharan
2011b). Such a component to credit availability could in fact be independent of economic fundamentals, expectations of future fundamentals, and
credit demand.
Using observations on credit standards, capital flows, and interest rates
for the United States and for a panel of eleven countries, we present evidence
on how these variables are related to real house price movements in recent
data. Our main measure of credit standards is compiled from quarterly bank
surveys of senior loan officers, carried out by national central banks as part
of their regulatory oversight. We consider this a summary indicator of fluctuations in the variables associated with an FML, as described earlier. The
surveys specifically address changes in a bank’s supply of credit, as distinct
from changes in its perceived demand for credit. We find for the United
States that this measure of credit supply, by itself, explains 53 percent of the
quarterly variation in house price growth over the period 1992 to 2010, while
it explains 66 percent over the period since 2000. By contrast, controlling for
credit supply, various measures of capital flows, real interest rates, and aggregate activity (collectively) add less than 5 percent to the fraction of variation
explained for these same movements in home values. Credit supply retains
its strong marginal explanatory power for house price movements over the
period 2002 to 2010 in a panel of international data, while capital flows have
no explanatory power. Moreover, credit standards continues to be the most
important variable related to future home price fluctuations even when it
has been rendered statistically orthogonal to banks’ perceptions of credit
demand, and even when controlling for expected future economic growth
and expected future real interest rates. Taken together, these findings suggest
that a stark shift in bank lending practices—conspicuous in the FML and
its reversal—was at the root of the housing crisis.
240
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
The rest of this chapter is organized as follows. The next section discusses theoretical literature that has addressed the link between house prices,
capital flows, and/or credit supply. To provide a theoretical frame of reference, here we also describe in detail the predictions of FLVNa for house
price movements. Section 6.3 turns to the data, presenting stylized facts
on international capital flows, interest rates, and credit standards. Section
6.4 presents an empirical analysis of the linkage between capital flows and
house price fluctuations, controlling for measures of credit supply, economic
activity, and real interest rates. Section 6.5 concludes. The appendix provides
details on the data we use and on our estimation methodology.
6.2
Theories
A number of studies have addressed the link between house prices and
capital flows, focusing on the recent boom period in housing. For brevity,
we will refer to the period of rapid home price appreciation from 2000 to
2006 as the boom period in the United States, and the period 2007 to present as the bust.
The global savings glut hypothesis (Bernanke 2005; Mendoza, Quadrini,
and Rios-Rull 2007; Bernanke 2008; Caballero, Farhi, and Gourinchas
2008; Caballero and Krishnamurthy 2009) contends that a number of possible events (the Asian financial crisis in the late 1990s being one frequently
cited) led to an increase in savings in developing countries, notably China
and emerging Asia, which sought safe, high- quality financial assets that
their own economies could not provide. Because of the depth, breadth, and
safety of US Treasury and Agency markets, those savings predominantly
found their way to the United States. To the extent that saving in developed
nations remained roughly unchanged by these events, the increase in savings in developing nations would cause an increase in worldwide savings,
hence the global savings glut. Some have directly linked these capital flow
patterns to higher US home prices, arguing that low interest rates (driven
in part by the capital inflow) were a key determinant of higher house prices
during the boom (e.g., Bernanke 2005; Himmelberg, Mayer, and Sinai 2005;
Bernanke 2008; Taylor 2009; Adam, Marcet, and Kuang 2011). In a similar
spirit, Caballero and Krishnamurthy (2009) identify the start of the housing boom with the Asian financial crisis, which fueled the demand for US
risk- free assets. In their model, Asian savers turn to US assets, resulting in
a net capital inflow for the United States. Global interest rates then fall in
their model because the US economy is presumed to grow more slowly than
the rest of the world.
Laibson and Mollerstrom (2010) have criticized the global savings glut
hypothesis by noting that an increase in worldwide savings should have led
to an investment boom in countries that were large importers of capital,
notably the United States. Instead, the United States experienced a con-
International Capital Flows and House Prices
241
sumption boom that accompanied the housing boom, suggesting that saving
worldwide was not unusually high. Laibson and Mollerstrom (2010) present
an alternative interpretation of the correlation between home values and
capital flows during the boom based on asset bubbles. Assuming a bubble in
the housing market, they argue that the rise in housing wealth generated by
the bubble led to higher consumption, which in turn led to greater borrowing from abroad and a substantial net capital inflow to the United States. A
similar idea is presented in Ferrero (2011), but without the bubble. Ferrero
studies a two- sector representative- agent model of international trade in
which lower collateral requirements facilitate access to external funding and
drive up house prices.
Others have argued that preference shocks and a desire for smooth (across
goods) consumption can generate a correlation between house prices and
capital inflows. Gete (2010) shows that consumption smoothing across tradable (nonhousing) goods and nontradable (housing) goods can lead to a
positive correlation between house prices and current account deficits. With
an exogenous increase in the home country’s preference for housing, productive inputs in the home country are reallocated toward housing production,
so that housing consumption can rise. But with a preference for smooth
consumption across goods, the tradable nonhousing good (presumed identical across countries) will then be imported from abroad, leading to capital
inflows to the home country.
The abovementioned theories fall into two broad categories: those that
rely on higher domestic demand to drive both house prices and capital
inflows in the same direction (Gete 2010; Laibson and Mollerstrom 2010;
Ferrero 2011), and those that rely on capital inflow–driven low interest rates
to drive up house prices (Bernanke 2005; Himmelberg, Mayer, and Sinai
2005; Bernanke 2008; Taylor 2009; Caballero and Krishnamurthy 2009;
Adam, Marcet, and Kuang 2011). While these papers were motivated by
observations on housing and capital flows during the housing boom, they
also have implications for the housing bust. The former imply that the housing bust should be associated with a reversal of domestic demand, leading to
a capital outflow. The latter imply that the housing bust should be associated
with a rise in real interest rates, driven by a capital outflow.
As we show in the following, recent data pose a number of challenges to
these theories. First, while it is true that real interest rates were low throughout the boom period, they have remained low and have even fallen further
in the bust period. Second, while capital certainly flowed into countries such
as the United States during the boom period, there is no evidence of a clear
reversal in this trend during the bust period.3 These observations suggest
3. Some empirical studies document a positive correlation between house prices and capital
inflows to the United States, but these studies typically have data samples that terminate at the
end of the boom or shortly thereafter (e.g., Aizenman and Jinjarak 2009; Kole and Martin
2009).
242
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
that the economic and political forces responsible for driving capital flows
and house prices over the entire period were, to a large extent, distinct. Later
we present empirical evidence that neither capital inflows nor real interest
rates bear a strong relation to house prices in a sample that includes both
the boom and the bust.
We interpret these recent events through the lens of the theoretical models in FLVNa and FLVNb, focusing specifically on the model in FLVNa in
which an FML and its reversal are studied. Rather than reproducing the
mathematical description of the model here, we simply describe it verbally
and refer the reader to the original papers for details. Our focus here is on
empirical evidence relating home prices to various indicators as a means of
distinguishing among theories. Next we describe the model in FLVNa, and
explain how it differs from the abovementioned theories.
6.2.1
The Housing Boom-Bust: A Theory of Time-Varying Risk Premia
The FLVNa paper studies a two- sector general equilibrium model of
housing and nonhousing production where heterogenous households face
limited risk- sharing opportunities as a result of incomplete financial markets. A house in the model is a residential durable asset that provides utility to the household, is illiquid (expensive to trade), and can be used as
collateral in debt obligations. The model economy is populated by a large
number of overlapping generations of households who receive utility from
both housing and nonhousing consumption and who face a stochastic life
cycle earnings profile. We introduce market incompleteness by modeling
heterogeneous agents who face idiosyncratic and aggregate risks against
which they cannot perfectly insure, and by imposing collateralized borrowing constraints on households.
Within the context of this model, FLVNa focus on the macroeconomic
consequences of three systemic changes in housing finance, with an emphasis on how these factors affect risk premia in housing markets, and how risk
premia in turn affect home prices. First, FLVNa investigate the impact of
changes in housing collateral requirements.4 Second, they investigate the
impact of changes in housing transactions costs. Taken together, these two
factors represent the theoretical counterpart to the real- world FML discussed before. Third, FLVNa investigate the impact of an influx of foreign
capital into the domestic bond market. They argue that all three factors
fluctuate over time and changed markedly during and preceding the period
4. Ferrero (2011) also assumes a relaxation of credit constraints to explain the housing boom.
A key distinction between his model and FLVNa, however, is that Ferrero studies a two- country
representative agent model, so an increase in borrowing by the domestic agent is only possible
with an increase in lending from the rest of the world, hence a higher current account deficit. By
contrast, in FLVNa, borrowing and lending can happen within the domestic economy between
heterogeneous agents, so housing finance need not be tied to foreign savings. Thus, a reversal
of the FML in a setting like that of Ferrero’s would necessitate a capital outflow, whereas in
FLVNa it does not.
International Capital Flows and House Prices
243
of rapid home price appreciation from 2000 to 2006, and the subsequent
bust. In particular, the boom period was marked by a widespread relaxation of collateralized borrowing constraints and declining housing transactions costs (including costs associated with mortgage borrowing, home
equity extraction, and refinance). The period was also marked by a sustained
depression of long- term interest rates that coincided with a vast inflow of
capital into US safe bond markets. In the aftermath of the credit crisis that
began in 2007, the erosion in credit standards and transactions costs has
been sharply reversed.5 We provide evidence on this following.
The main impetus for rising price- rent ratios in the model in the boom
period is the simultaneous occurrence of positive economic shocks and a
financial market liberalization, phenomena that generate an endogenous
decline in risk premia on housing and equity assets. As risk premia fall,
the aggregate house price index relative to aggregate rent rises. An FML
reduces risk premia for two reasons, both of which are related to the ability
of heterogeneous households to insure against aggregate and idiosyncratic
risks. First, lower collateral requirements directly increase access to credit,
which acts as a buffer against unexpected income declines. Second, lower
transactions costs reduce the expense of obtaining the collateral required
to increase borrowing capacity and provide insurance. These factors lead to
an increase in risk- sharing, or a decrease in the cross- sectional variance of
marginal utility. The housing bust is caused by a reversal of the FML and
of the positive economic shocks and an endogenous decrease in borrowing
capacity as collateral values fall. These factors lead to an accompanying rise
in housing risk premia, driving the house price- rent ratio lower. Almost all
of the theories discussed previously are silent on the role of housing risk
premia in driving house price fluctuations.6
It is important to note that the rise in price- rent ratios caused by a financial market liberalization in FLVNa must be attributed to a decline in risk
premia and not to a fall in interest rates. Indeed, the very changes in housing
finance that accompany a financial market liberalization drive the endogenous interest rate up, rather than down. It follows that, if price- rent ratios
rise after a financial market liberalization, it must be because the decline in
risk premia more than offsets the rise in equilibrium interest rates that is
attributable to the FML. This aspect of an FML underscores the importance of accounting properly for the role of foreign capital over the housing
cycle. Without an infusion of foreign capital, any period of looser collateral
requirements and lower housing transactions costs (such as that which characterized the housing boom) would be accompanied by an increase in equilibrium interest rates, as households endogenously respond to the improved
5. Streitfeld (2009) argues that, since the credit crisis, borrowing restrictions and credit constraints have become even more stringent than historical norms in the preboom period.
6. An exception is Caballero and Krishnamurthy (2009), but they do not study housing nor
the FML and its reversal.
244
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
risk- sharing opportunities afforded by a financial market liberalization by
reducing precautionary saving.
To model capital inflows, FLVNa introduce foreign demand for the domestic riskless bond into the market- clearing condition. This foreign capital
inflow is modeled as driven by governmental holders who inelastically place
all of their funds in domestic riskless bonds. Foreign governmental holders
have a perfectly inelastic demand for safe securities and place all of their
funds in those securities, regardless of their price relative to other assets.
Later we discuss data on US international capital flows that supports this
specification of the net capital flows in the United States over the last fifteen
years.
The model in FLVNa implies that a rise in foreign purchases of domestic
bonds, equal in magnitude to those observed in the data from 2000 to 2010,
leads to a quantitatively large decline in the equilibrium real interest rate.
Were this decline not accompanied by other, general equilibrium effects,
it would lead to a significant housing boom in the model. But the general
equilibrium effects imply that a capital inflow is unlikely to have a large
effect on house prices even if it has a large effect on interest rates. One reason
for this involves the central role of time- varying housing risk premia. In
models where risk premia are held fixed, a decline in the interest rate of this
magnitude would be sufficient—by itself—to explain the rise in price- rent
ratios observed from 2000 to 2006 under reasonable calibrations. But with
time- varying housing risk premia, the result can be quite different. Foreign
purchases of US bonds crowd domestic savers out of the safe bond market,
exposing them to greater systematic risk in equity and housing markets.
In response, risk premia on housing and equity assets rise, substantially
offsetting the effect of lower interest rates and limiting the impact of foreign
capital inflows on home prices.
There is a second offsetting general equilibrium effect. Foreign capital
inflows also stimulate residential investment, raising the expected stock of
future housing and lowering the expected future rental growth rate. Like risk
premia, these expectations are reflected immediately in house prices (pushing down the national house price- rent ratio), further limiting the impact of
foreign capital inflows on home prices. The net effect of all of these factors
is that a large capital inflow into safe securities has, at most, a small positive
effect on house prices.
It is useful to clarify the two opposing forces simultaneously acting on
housing risk premia in the model of FLVNa. During the housing boom,
there is both an FML and a capital inflow. As explained, the FML lowers risk
premia, while foreign purchases of domestic safe assets raise risk premia.
Under the calibration of the model, the decline in risk premia resulting from
the FML during the boom period is far greater than the rise in risk premia
resulting from the capital inflow. On the whole, therefore, risk premia on
housing assets fall, and this is the most important contributing factor to the
International Capital Flows and House Prices
245
increase in price- rent ratios during the boom. During the bust, modeled as
a reversal of the FML but not the capital inflows, risk premia unambiguously rise even as interest rates remain low. The rise in risk premia drives the
decline in house- price rent ratios.
These features of the model represent significant differences from other
theories of capital flows and house prices. They permit the model to explain
not just the housing boom, but also the housing bust, in which house pricerent ratios fell dramatically, even though interest rates remained low and
there has been no clear reversal in the trend toward capital inflows into the
US bond market. Moreover, they underscore the importance of distinguishing between interest rate changes (which are endogenous) and credit supply.
In the absence of a capital inflow, an expansion of credit supply in the form
of lower collateral requirements and lower transactions costs should lead,
in equilibrium, to higher interest rates, rather than lower, as households
respond to the improved risk- sharing/insurance opportunities by reducing
precautionary saving. Instead we observed low real interest rates, generated
in the model of FLVNa by foreign capital inflows, but the inflows themselves
are not the key factor behind the housing boom- bust.
To illustrate the independent role of house prices and capital inflows in
the model, figure 6.2 plots the transition dynamics for both the aggregate
price- rent ratio and for foreign holdings of domestic assets over the period
2000 to 2010 from the model of FLVNa. The figure shows the dynamic
behavior of the price- rent ratio in response to a series of shocks designed
to mimic both the state of the economy and housing market conditions over
the period 2000 to 2010. The economy begins in year 2000 at the stochastic
steady state of a world with “normal” collateral requirements (i.e., fraction
of home value that must be held as collateral) and housing transactions costs
calibrated to roughly match the data prior to the housing boom of 2000 to
2006. In 2001, the economy undergoes an unanticipated shift to a new steady
state, in which there is an FML with lower collateral requirements and lower
transactions costs, calibrated to match the changes in these variables during
the boom period, as well as an unanticipated increase in foreign holdings of
US bonds from 0 to 16 percent of GDP. This 16 percent increase is calibrated
to match the actual increase in net foreign holdings of US securities over the
period 2000 to 2010. Along the transition path, foreign holdings of bonds
are increased linearly from 0 to 16 percent of GDP from 2000 to 2010. The
adjustment to the new stochastic steady state is then traced out over the
seven- year period from 2001 to 2006, as the state variables evolve. Finally,
starting in 2007 and continuing through 2010, the economy is presumed to
undergo a surprise reversal of the financial market liberalization but not the
foreign capital inflow.
Figure 6.2 shows that the house- price rent ratio rises by 39 percent over
the period 2000 to 2006 and then falls by 17 percent over the period 2006 to
2010. By contrast, foreign holdings of domestic riskless bonds, denoted BFt ,
246
Fig. 6.2
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
Price-rent ratio and foreign holdings in FLVNa
Source: Favilukis, Ludvigson, and Van Nieuwerburgh (2010).
Notes: The figure plots the transition dynamics of the aggregate price- rent ratio and foreign holdings of the safe asset in the model of Favilukis, Ludvigson, and Van Nieuwerburgh
(2010), for the period 2000 to 2010. The dynamics are driven by a sequence of aggregate
economic shocks designed to mimic the business cycle over this period, a financial market
liberalization in 2000, and a reversal of this liberalization in 2007, as well as the foreign flows
depicted in the figure.
rise at a constant rate throughout the boom- bust period. Although foreign
holdings rise mechanically over time and are crudely calibrated to match the
long- term (trend) increase in holdings over the entire ten- year period (rather
than matching the year- by- year fluctuations), the figure nevertheless shows
that capital flows are not a key determinant of the boom- bust pattern in the
price- rent ratio in this model, despite the large decline in interest rates generated by these inflows. In the data, the increase in the price- rent ratio (series
shown in figure 6.1) over the period 2000:Q4 to 2006:Q4 is 49.9 percent
(calculated in the same way as in the model), while over the bust (2006:Q4
to 2010:Q4) it declined 34.0 percent. The model captures 78 percent of the
run-up in this measure and 49 percent of the decline.
The relationship between capital inflows and risk premia in FLVNa and
FLVNb is worthy of emphasis. In equilibrium, higher capital inflows into
the safe bond market raise risk premia on housing and equity, rather than
International Capital Flows and House Prices
247
lower them. This runs contrary to the argument, made by some, that the
free flow of capital across borders should be associated with a reduction in
risk premia (e.g., Geithner 2007). Here, foreign purchases of the safe asset
make both equity and housing assets more risky. Both the risk premium and
Sharpe ratio for equity and housing rise when there is a capital inflow, for
two reasons. First, the increase in foreign money forces domestic residents
as a whole to take a leveraged position in the risky assets. This by itself
increases the volatility of asset and housing returns, translating into higher
risk premia. Second, domestic savers are crowded out of the bond market
by foreign governmental holders who are willing to hold the safe asset at
any price. As a result, they become more exposed to systematic risk in the
equity and housing markets. This means that the equity and housing Sharpe
ratios must rise, as domestic savers shift the composition of their financial
wealth toward risky securities. In addition, the volatility of the stochastic
discount factor rises and there is a decrease in risk- sharing, as measured by
the cross- sectional variance of marginal utility growth.
Of course, the effect of a capital inflow on house prices depends not
only on the housing risk premium, but also on the risk- free interest rate.
Although a capital inflow drives the housing risk premium up, in the model
of FLVNa it drives the risk- free rate down by more, so a capital inflow still
leads to a modest increase in the price- rent ratio.7 In this model, an inflow
of foreign capital calibrated to match the increase in foreign ownership of
US Treasuries and US agency debt over the period 2000 to 2010 has a large
downward impact on the equilibrium interest rate, which falls from 3.45 percent to 0.39 percent. The magnitude of this decline is close to the reduction
in real rates observed in US bond market data over the period 2000 to 2006.
With this discussion as theoretical background, we now turn to an analysis of the data on capital flows, interest rates, and credit standards over the
boom- bust period.
6.3
Trends in Capital Flows, Interest Rates, Credit Supply
While the notion of a global savings glut is controversial, recent data
clearly suggest a reallocation of savings away from the developed world, and
toward the developing world, the so-called global imbalances phenomenon.
Unlike any prior period, global financial integration allowed for the channeling of one country’s excess savings toward another country’s real estate
boom. Such financing occurred directly, for example, by German banks’
purchases of US subprime securities, but also indirectly through the US
Treasury and Agency bond markets. As the world’s sole supplier of a global
7. Changes in expected future aggregate rent growth also can effect the price- rent ratio. The
numbers here refer to a comparison of stochastic steady states, however, in which the expected
rental growth rate is the same in both steady states (equal to the deterministic growth rate of
the economy).
248
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
reserve currency, the US experienced a surge in foreign ownership of US
Treasuries and Agency bonds. Agency bonds refers to the debt of the two
government- sponsored enterprises (GSEs) Freddie Mac and Fannie Mae, as
well as to the mortgage- backed securities that they issue and guarantee. Due
to their ambivalent private- public structure and their history as agencies of
the federal government, private market investors (including foreign investors) have always assumed that the debt of Freddie Mac and Fannie Mae
was implicitly backed by the US Treasury. That implicit backing became an
explicit backing in September 2008 when Freddie Mac and Fannie Mae were
taken into government conservatorship. See Acharya, Richardson, et al.
(2011) for details on the GSEs.
In this section, we discuss in detail data showing the trends in capital
flows, US real interest rates, and the relaxation and subsequent tightening
of housing credit constraints and credit standards.
6.3.1
International Capital Flows
The Treasury International Capital (TIC) reporting system is the official
source of US securities flows data. It reports monthly data (with a six- week
lag) on foreigners’ purchases and sales of all types of financial securities
(equities, corporate, Agency, and Treasury bonds). We refer to these monthly
transactions data as the TIC flows data. The TIC system also produces periodic benchmark surveys of the market value of foreigners’ net holdings, or
net asset positions, in US securities. Unlike the flows data, these data take
into account the net capital gains on gross foreign assets and liabilities. We
refer to these as the TIC holdings data. The holdings data are collected in
detailed surveys conducted in December of 1978, 1984, 1989, and 1994,
in March 2000, and annually in June from 2002 to 2010. The survey data
on holdings is thought to be of higher quality than the flows data because
it more accurately accounts for valuation effects (Warnock and Warnock
2009).8
The Bureau of Economic Analysis (BEA) in the US Department of Commerce also provides annual estimates of the value of accumulated stocks
(holdings) of US- owned assets abroad and of foreign- owned assets in the
United States. We will refer to these as the BEA holdings data. These include
estimates of holdings of securities, based on the TIC data, as well as estimates of holdings of other assets such as foreign direct investment (FDI),
US official reserves, and other US government reserves. We refer to the sum
of these other assets plus financial securities as total assets. In recent data,
the main difference between the BEA estimate of net foreign holdings of
total assets and its estimate of net foreign holdings of total securities is
8. As explained in Warnock and Warnock (2009), reporting to the surveys is mandatory, with
penalties for noncompliance, and the data are subjected to extensive analysis and editing. Data
on foreign holdings of US securities are available at http://www.treasury.gov/resource- center
/data- chart- center/tic/Pages/index.aspx.
International Capital Flows and House Prices
249
attributable to FDI, where, since 2006, the value of US FDI abroad has
exceeded the value of foreign FDI in the United States.9
The BEA defines the US net international investment position (NIIP) as the
value of US- owned assets abroad minus foreign- owned assets in the United
States. The overall change in the NIIP incorporates capital gains and losses
on the prior stock of holdings of assets. Thus, the total change in US gross
foreign assets equals net purchases by US residents plus any capital gains on
the prior stock of gross foreign assets, while the total change in US foreign
liabilities equals net sales of assets to foreign residents plus any capital gains
accrued to foreigners on their US assets. The change in the NIIP is the difference between the two. Capital gains are the most important component of
valuation changes on the NIIP.
The BEA also collects quarterly and annual estimates of transactions with
foreigners, including trade in goods and services, receipts and payments of
income, transfers, and transactions in financial assets. We refer to these as the
BEA transactions data. The transactions data measure the current account
(CA). Since the CA transactions data only measure purchases and sales
of assets, they do not adjust for valuation effects that must be taken into
account in constructing the international investment positions (holdings) of
the United States, as just discussed.10
When thinking about the recent boom- bust period in residential real
estate, a question arises as to which measure of capital flows to study. Obstfeld (2011) documents an increase in the sheer volume of financial trade
across borders, and argues that it could be positively correlated with financial instability. Moreover, he shows that the amplitude of pure valuation
changes in the NIIP has grown in tandem with the volumes of gross flows.
Because the CA ignores such valuation changes, our preferred measure
would therefore be a measure of total changes in net foreign holdings of
assets rather than changes in net transactions. Unfortunately, data on net
foreign asset holdings are only readily available in the United States, and
then only annually. (For the empirical work following, we construct our
own quarterly estimate of these holdings for securities.) Outside the United
States, only the transactions- based CA data are available. Thus, when we
use international data we use the CA as a measure of capital flows, bearing
in mind the limitations of these data for measuring changes in actual asset
holdings.
Since net foreign asset holdings data are available for the United States,
when working with US data we focus most on net foreign holdings as a
measure of capital flows (although for completeness we also present empirical results using the CA as a measure of capital flows). Within net foreign
9. These data are available at http://www.bea.gov/international/index.htm.
10. See the adjustments for valuations effects at http://www.bea.gov/international/xls
/intinv10_t3.xls.
250
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
holdings, we focus on changes in holdings of financial securities, rather
than changes in holdings of total assets. We argue that the former are far
more relevant for residential real estate than the latter. Recall that the most
important difference between the two, especfially in recent data, is attributable to flows in FDI. But it is unclear how relevant FDI is for the housing
market. For example, during much of the housing boom, the value of net
foreign holdings on FDI fell, implying a net capital outflow on those types
of assets. This fact is hardly consistent with the notion that capital inflows
to the United States helped finance the housing boom.
What flowed in during the housing boom was foreign capital directed at
US Treasury and Agency securities. There are several reasons we expect these
assets—unlike FDI—to be directly related to the US housing market. First,
foreign purchases of Agency securities allowed the government- sponsored
enterprises (GSEs) Fannie Mae and Freddie Mac to broaden their market
for mortgage- backed securities (MBS) to international investors, funding
the mortgage investments themselves. Thus, an inflow of capital into US
Agencies can in turn free up US banks to fund additional mortgages. Second, because mortgage rates are often tied to Treasury rates, large foreign
Treasury purchases could in principle directly affect house prices through
their effect on interest rates. In addition, low Treasury rates could lead US
banks in search of yield to undertake more risky mortgage investments (see
Maddaloni and Peydro [2011] for evidence that banks increase the riskiness
of investments in low interest rate environments). In summary, because the
FDI streams are largely divorced from the US housing market, the most
appropriate measure of capital flows for our purpose is not net foreign holdings of total assets but instead total securities.
Figure 6.3 shows the movement in various measures of international
capital flows into the United States, relative to trend GDP, in annual data
from 1976 to 2010. Plotted are the change in net foreign holdings of total
assets, total securities, and in what we will call US “safe” securities (defined
as Treasuries and Agencies). We refer to a capital inflow as a positive change
in holdings, and vice versa for a capital outflow. Also plotted is the current
account deficit. Figure 6.3 shows that there is considerable volatility in these
measures during the housing boom and the subsequent financial crisis, with
particularly sharp increases in the change in net foreign holdings of US assets
from 2007 to 2008. This corresponds to an upward spike in the change in the
US net foreign liability position in 2008 (change in net foreign holdings of
total assets in the figure). This series declines from 2008 to 2009 and increases
again from 2009 to 2010. Comparing the endpoints of these series in 2010 to
their values in 2006, we see that—by any measure of assets—inflows (or the
change in holdings) were higher at the end of the sample in 2010 than they
were at the peak of the housing boom at the beginning of 2006 (end of 2005).
To get a better sense of the trends in these series, figure 6.4 plots the same
measures of international capital flows, but computed as four- year moving
Fig. 6.3
Measures of US capital flows
Source: US Department of Commerce, Bureau of Economic Analysis.
Notes: Net foreign holdings of total assets is foreign- owned assets in the United States minus
US- owned assets abroad. Net foreign holdings of total securities is defined as foreign- owned
US government securities plus US Treasury securities plus US securities other than Treasury
securities minus US- owned foreign securities. Net foreign holdings of safe securities is defined
as US government securities plus Treasury securities. (We do not subtract off US holdings
of foreign government securities, since these carry at least exchange- rate risk.) The current
account deficit is imports of goods and services and income payments less exports of goods and
services and income receipts less net unilateral current transfers. All series are expressed relative
to trend GDP (obtained from a Hodrick–Prescott trend). The sample is annual, 1976 to 2010.
Fig. 6.4
Measures of US capital flows, four-year moving average
Source: US Department of Commerce, Bureau of Economic Analysis.
Notes: See figure 6.3. The sample is annual, 1980 to 2010.
252
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
averages. The figure shows that changes in net foreign holdings of total assets
show little trend movement over the entire boom- bust period 2000 to 2010,
but if anything they trended up during the bust period from 2006 to 2010,
while they trended down in the boom from 2002 to 2006. A similar pattern
holds for net foreign holdings of total securities, except that here, inflows are
much more sharply positive during the housing bust period. So where are
the inflows during the housing boom? In US safe securities.
The only assets for which we observe a significant increase in capital
inflows during the boom period are those we defined earlier as US safe
securities, comprised of Treasury and Agency debt. We have argued earlier
that these assets are likely to be the most relevant for housing markets, and
indeed the change in net foreign holdings of these securities was positive and
increasing throughout the boom period, from 2001 through the beginning
of 2006, which we take as the peak of the housing boom. At the same time,
however, inflows into these securities, like the other categories of assets,
continued to rise during the bust period, implying that the US borrowing
from abroad in these securities increased further from the beginning of 2006
to end 2010, rather than declined.
The only measure in figure 6.4 that suggests a decline in the rate at which
the United States is borrowing from abroad is the current account deficit,
pointing to a significant incongruence with the holdings data. (We discuss
this further later.) Despite the volatility in the holdings data, the bust period
still exhibited relatively high average inflows of foreign capital into all forms
of US securities, mirroring capital gains on US Liabilities relative to US
assets abroad, as well as a net flight into US safe securities, in 2008. The CA
deficit (or, equivalently, the capital account) omits these significant valuation
changes during the financial crisis. We view this as a serious shortcoming
of the CA as a measure of international capital flows, since such valuation
adjustments surely have wealth effects that in general equilibrium would influence the extent to which US households can consume at rates that exceed
domestic income. At the end of the sample in 2010, figure 6.4 shows that
there is a decline in the (moving average trend) inflows to total assets from the
end of 2008 to the end of 2010. But this decline is barely discernible in total
securities and is not at all present in US safe securities. The discrepancy is
again net flows into FDI, which we have already argued are largely divorced
from the housing market.
How can we reconcile the large decline in the current account deficit from
the end of 2005 to the end of 2010, with the observation that the change
in net foreign holdings of total US assets rose over this period (figure 6.3)?
Comparing 2010 to 2005 (year end), the current account deficit fell by
$274,876 million, while the year- end change in net foreign holdings of total
US assets (relative to trend GDP) rose by $395,440 million. The discrepancy is attributable to valuation effects, which the current account ignores.
Indeed, 126 percent of the discrepancy over this period is attributable to
International Capital Flows and House Prices
253
valuation effects (−26 percent is attributable to a statistical discrepancy and
other small adjustments between the current and capital account flows).
Thus, the decline in the current account deficit from 2005 to 2010 suggests a
decline in the rate at which US liabilities are increasing, when in fact this rate
has increased, primarily because the change in capital gains foreign residents
enjoyed on US assets from 2005 to 2010 far exceeded the change in capital
gains accruing to US residents on their assets abroad. But these valuation
adjustments came primarily from assets other than what we have defined as
US safe assets. (This is perhaps not surprising since these assets are far less
volatile than is risky capital.) A breakdown suggests that only 15.6 percent
of these valuation adjustments (specifically of the change in these adjustments from 2005 to 2010) came from adjustments on US safe assets. A much
larger 39.7 percent came from financial securities other than safe securities,
and the majority (44.7 percent) came from valuation adjustments on assets
other than financial securities (including both safe and nonsafe financial
securities).
We can also compute the fraction of the cumulative change in net foreign
holdings of safe assets from the end of 2005 to the end of 2010 that is attributable to valuation changes versus transactions. Over this period, transactions account for 92.6 percent, while valuation changes account for just
7.3 percent. This shows that, even accumulating over the entire bust period,
there continues to be a strong inflow of capital into US safe securities that
is not attributable merely to valuation changes.
To summarize, during the housing boom, only US capital inflows on securities (equities, corporate, US Agency and Treasury bills and bonds) show
any discernible upward trend. Among securities, the upward trend has been
driven almost entirely by an increase in net foreign holdings of US safe
assets, specifically US Treasury and Agency debt. Yet net inflows on these
securities, rather than declining during the housing bust, have continued to
increase.
We now provide more detail on the flows to US safe securities. To get a
better sense of the quantitative importance of these flows to US safe assets,
the solid line of figure 6.5, measured against the left axis, plots the combined
foreign holdings in billions of US dollars of short- term and long- term US
Treasuries and Agencies. The dashed line, measured against the right axis,
shows long- term (not short- term) foreign holdings of Treasuries and Agencies relative to the amount of long- term marketable debt outstanding. Figure 6.6 plots total foreign holdings of US Treasuries and Agencies relative
to the size of the US economy, measured as trend GDP.
The figure shows that foreign holdings of US Treasuries were modest until
the mid- 1990s. In December 1994, foreign holdings of long- term Treasuries
were $464 billion, which amounted to 19.4 percent of marketable Treasuries outstanding and to 6.4 percent of US trend GDP. Foreign holdings
of long- term Agencies were $121 billion, which amounted to 5.4 percent of
254
Fig. 6.5
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
Foreign holdings of US safe assets
Source: US Treasury International Capital System’s annual survey of foreign portfolio holdings of US securities.
Notes: The figure plots foreign holdings of US Treasuries and US Agencies (circles). US Agencies denotes both the corporate bonds issued by the government- sponsored enterprises and
the mortgage- backed securities guaranteed by them. The solid lines denote the amount of
long- term and short- term holdings, in billions of US dollars, as measured against the left axis.
The dashed lines denote the long- term foreign holdings relative to the total amount of outstanding long- term (marketable) debt. The foreign holdings data are available for December
1974, 1978, 1984, 1989, 1994, 1997, March 2000, and annually from June 2001 until June 2010.
outstanding Agencies and 1.5 percent of trend GDP. Over the course of the
Asian financial crisis, these holdings doubled. By March 2000, toward the
end of the crisis, foreign holdings of long- term Treasuries and Agencies were
$884 billion and $261 billion, respectively, corresponding to 35.3 percent and
7.3 percent of the amounts outstanding. Total foreign holdings of Treasuries
and Agencies increased from 9.8 percent to 14.8 percent of trend GDP.
Caballero, Farhi, and Gourinchas (2008) argue that the Asian financial crisis
represented a negative shock to the supply of (investable/pledgeable) assets
in East Asia, and led their investors to increase their investments in US
bonds, one of the scarce risk- free assets available worldwide.
During the housing boom from 2000 to 2006, the increase in foreign holdings of safe assets continued at an even more rapid pace. Total foreign holdings of Treasuries and Agencies more than doubled, from $1,418 billion in
March 2000 to $3,112 billion in June 2006. Foreign holdings of long- term
Treasuries went from 35.3 percent to 52.0 percent of the total amount of
Treasuries outstanding, while holdings of long- term Agencies went from
7.3 percent to 17.2 percent. Most of the rise in foreign holdings of Treasuries
took place by 2004, while most of the rise in Agencies took place from 2004
International Capital Flows and House Prices
Fig. 6.6
255
Foreign holdings relative to GDP
Source: US Treasury International Capital System’s annual survey of foreign portfolio holdings of US securities.
Notes: The solid line denotes foreign holdings of US Treasuries and Agencies relative to US
trend GDP (squares). Trend GDP is computed with a Hodrick–Prescott filter (Hodrick and
Prescott 1997). The dashed line (circles) asks what the foreign holdings relative to trend GDP
would have been if the foreign holdings relative to the amount of debt outstanding declined
the amount they did but the amount of debt outstanding relative to trend GDP was held at
2008 values for the years 2009 and 2010. The foreign holdings data are available for December
1974, 1978, 1984, 1989, 1994, 1997, March 2000, and annually from June 2001 until June 2010.
to 2006. Total foreign holdings of Treasuries and Agencies increased from
14.8 percent to 23.7 percent of trend GDP. The boom in US house prices,
which started at the end of 1994 and accelerated after 2000, coincided with
a massive inflow of foreign capital in safe US assets.
At the same time, however, capital inflows in the US safe assets continued
to rise during the housing bust and financial crisis. Figure 6.5 shows that
between June 2006 and June 2010, total foreign holdings of Treasuries and
Agencies rose from $3,112 billion to $5,232 billion, or from 23.7 percent
to 35.5 percent of trend GDP. The share of outstanding long- term Treasuries held by foreign investors also increased from 52.0 percent in 2006
to 61.1 percent in 2008 before falling back to 53.0 percent in 2010. The
reduction in 2010 is attributable to a large increase in the total quantity
of marketable Treasuries outstanding in 2009 and 2010 (which rose from
33.2 percent of trend GDP in 2008 to 54.9 percent in 2010), rather than to
a reduction in nominal foreign holdings. The latter actually continued to
increase rapidly from $2,211 billion in 2008 to $3,343 billion in 2010. The
256
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
dashed line in figure 6.6 is a foreign holdings- to-trend GDP series that we
have adjusted in 2009 and 2010 to reflect the large increase in the quantity
of Treasury debt outstanding that occurred in 2009 and 2010. The adjusted
series equals the level of foreign holdings as a fraction of trend GDP that
would have occurred in 2009 and 2010 had Treasury debt outstanding as a
fraction of trend GDP been fixed at its 2008 level. The dashed line shows
that the increase in foreign holdings of US Treasuries in 2009 and 2010 is
less than proportional to the increase in outstanding Treasuries over those
years. In this relative sense, therefore, foreigners have become less willing to
hold US Treasuries. According to the adjusted series there is a reduction in
foreign holdings as a fraction of trend GDP, from 30.0 percent of trend GDP
in 2008, to 24.6 percent in 2010, suggesting that a substantial “unwind” of
foreign positions in US Treasuries may be under way, at least relative to the
total amount of US debt being issued.
Although there has so far been no reduction in nominal foreign holdings
of US Treasuries during the housing bust, the financial crisis did lead to a
substantial reduction in nominal foreign holdings of US Agencies. While
foreign holdings of Agencies still rose from 17.2 percent in 2006 to 20.8
percent of the amount outstanding in 2008, they fell back sharply to 15.6
percent of the amount outstanding in 2010, even as the amount outstanding
remained flat.
Foreign Official Holdings
An important aspect of recent patterns in international capital flows is
that foreign demand for US Treasury securities is dominated by Foreign
Official Institutions. Krishnamurthy and Vissing-Jorgensen (2007) find that
demand for US Treasury securities by governmental holders is extremely
inelastic, implying that when these holders receive funds to invest they buy
US Treasuries, regardless of their price. As explained in Kohn (2002), government entities have specific regulatory and reserve currency motives for
holding US Treasuries and face both legal and political restrictions on the
type of assets that can be held, forcing them into safe securities.
Data from the TIC system breaks out what share of foreign holdings of
US Treasuries is attributable to Foreign Official Institutions, which are government entities, mostly central banks. Foreign Official Institutions own the
vast majority of US Treasuries in recent data: in June 2010 Foreign Official
Institutions held 75 percent of all foreign holdings of US Treasuries. That
share has always been high and has risen from 58 percent in March 2000 to
75 percent in June 2010. Indeed, 75 percent represents a lower bound on the
fraction of such securities held by Foreign Official Institutions, since some
prominent foreign governments purchase US securities through offshore
centers and third- country intermediaries, purchases that would not be
attributed to Foreign Official entities by the TIC system—see Warnock and
Warnock (2009). Foreign Official Institutions also accounted for 64 percent
of the foreign holdings of Agencies in June 2010.
International Capital Flows and House Prices
Fig. 6.7
257
Foreign holdings by maturity and by country
Source: US Treasury International Capital System’s annual survey of foreign portfolio holdings of US securities.
Notes: The right panel shows the share of foreign holdings of US Treasuries and Agencies
for different maturity ranges, expressed in years. The data are for June 2009 (table 14a of the
2009 TICS report). Fifty- one percent of the Agency holdings are reported to be of maturity
twenty-five to thirty years. These are thirty- year mortgage- backed securities, which have a
nominal maturity of thirty years but an effective maturity of about seven years because of
prepayment. We reallocate these thirty- year Agency holdings across maturities so that they
have a weighted average maturity (WAM) of seven years. The resulting series for the maturity
of foreign holdings of US Agencies has a WAM of 6.4 years instead of the original 17.7 years.
Total long- term foreign holdings have a WAM of 5.2 years. Once we add the short- term
Treasury holdings, which are one year or less in maturity, the WAM drops to 4.6 years. The
left panel shows the foreign holdings as of December 2010 by country groups. China excludes
Hong Kong, which is part of “Rest of Asia.” Banking centers consist of the United Kingdom,
the Caribbean, Luxembourg, Belgium, and Ireland.
Asian central banks (China, Japan, Korea) have acquired massive US
dollar reserves in the process of stabilizing their exchange rate. The share
of foreign holdings is higher for long- term than for short- term securities.
The left panel of figure 6.7 shows the foreign holdings as of December
2010 by country groups. China excludes Hong Kong, which is part of “Rest
of Asia.” Banking centers consist of the United Kingdom, the Caribbean,
Luxembourg, Belgium, and Ireland. It is widely believed that China holds a
nontrivial fraction of its safe dollar assets through financial intermediaries
in the UK and in other banking centers (Warnock 2010). The graph then
suggests that as much as two- thirds of safe US assets is held by Asian countries. China (narrowly defined) held nearly $1,500 billion in Treasuries and
Agencies in June 2010; Japan held nearly $1,000 billion.
The right panel of figure 6.7 shows the share of foreign holdings of US
Treasuries and Agencies for different maturity ranges, expressed in years.
The data are for June 2009. As the caption explains, the maturity of the
Agency holdings is adjusted to account for the prepayment option embedded in mortgage- backed securities. Total long- term and short- term foreign
258
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
holdings have a weighted average maturity of 4.6 years. About a quarter of
foreign holdings have a maturity of one year or less. Fully half of all holdings
have a maturity below three years. This suggests that a substantial reduction in foreign holdings of US safe assets could occur over a relatively short
period without an outright fire- sale of long- term bonds, if current holders
simply stopped rolling over existing positions.
Longer-Term Trends in Net Foreign Holdings of Securities
We have emphasized the special relevance for the US housing boom- bust
cycle of US securities considered to be safe stores- of-value (i.e., US Treasury and Agency debt). But is worth emphasizing that, even over a longer
period of time, foreign holdings of these securities behave similarly to total
net foreign holdings of all securities. The reason is that foreign holdings
of US securities other than Treasuries and Agencies are roughly equal in
magnitude to US holdings of securities abroad. Figure 6.8 makes this point
visually; net foreign holdings of all securities other than US Treasury and
Agency debt as a fraction of US Trend GDP have hovered close to zero
since 1994, even as net foreign holdings of safe securities have soared. This
shows that all of the long- term upward trend in net foreign holdings of US
securities since 1994 has been the result of an upward trend in net foreign
Fig. 6.8
Net foreign holdings relative to US trend GDP
Source: US Treasury International Capital System’s annual survey of foreign portfolio holdings of US securities.
Notes: The solid line (squares) denotes total net foreign holdings of long- term securities (the
net foreign liability position of the United States) relative to US trend GDP. Net foreign holdings are defined as foreign holdings of US securities minus US holdings of foreign securities.
We define as net foreign holdings of safe securities to be foreign holdings of US Treasuries and
Agencies. (We do not subtract off US holdings of foreign government securities, since these
carry at least exchange- rate risk.) The dashed line (circles) denotes the thus constructed net
foreign holdings in safe securities, while the dotted line (diamonds) denotes the net foreign
holdings in all other securities. The data are available for December 1994, December 1997,
March 2000, and annually from June 2002 until June 2010.
International Capital Flows and House Prices
259
holdings of US safe securities; net foreign holdings of other securities are
almost exactly zero in June of 2010. Thus the long- term downward trend in
the US net international investment position is well described by the foreign
holdings of US safe assets. Of the safe assets foreigners hold, 70 percent (on
average) over the period 1994 to 2010 are held in US Treasuries. The large US
current account deficits in the boom period are, to a large extent, the mirror
image of the increase in foreign holdings of US safe assets.11 This is true
in the aggregate net flows to the United States, but also for China. China’s
cumulative current account surplus between 2003 and 2007 largely matches
up with its acquisitions of US Treasuries and Agencies (Bernanke 2011).
Risky Mortgage Holdings
Although net flows into securities other than Treasuries and Agencies
have hovered around zero, there were substantial gross flows across borders
into private- label products such as mortgage- backed securities (MBS), collateralized debt obligations (CDO), and credit default swaps (CDS) with
nonprime residential or commercial real estate as the underlying or as the
reference entity. Because an average of 80 percent of such private- label MBS
principal received a AAA rating from the credit ratings agencies and earned
yields above those of Treasuries (see US Treasury Department 2011), large
foreign (as well as domestic) institutional investors were able and willing
to hold these assets on their books. The TICS data indicate that foreigners
held $594 billion of nonagency mortgage- backed securities in June 2007. By
June 2009, these holdings more than halved to $266 billion, after which they
stabilized at $257 billion in June 2010. Less than 10 percent of these are held
by foreign official institutions (US Treasury Department 2011).
Bernanke (2011) shows interesting cross- country differences in the composition of countries’ US investment portfolio. China and emerging Asia
held three- quarters of their US investments in the form of Treasuries and
Agencies in 2007. Their share of all AAA- rated securities was 77.5 percent,
while the AAA- rated share of all US securities outstanding was only 36 percent. European (as well as domestic) investors held only about one- third of
their US portfolio in the form of AAA- rated assets. Not only did Europeans
invest in non-AAA corporate debt, they accumulated $500 billion in US
asset- backed (largely mortgage- backed) securities between 2003 and 2007.
In addition to their different risk profiles over the housing cycle, Europe
and Asia differ by their current account positions. While the Asian economies ran a large current account surplus, financing the purchases of US assets
with large trade surpluses, Europe had a balanced current account over this
period. It financed the purchases of risky US assets by issuing external liabilities, mostly equity, sovereign debt, and asset- backed commercial paper
11. Though, as discussed earlier, an important discrepancy between the current account data,
based on transactions, and the net foreign assets holdings data, is that the former do not fully
adjust for valuation effects that are captured in the international holdings data.
260
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
(ABCP). A prototypical example of European holdings were AAA- rated
tranches of subprime MBS held by large banks through lightly- regulated
off- balance sheet vehicles, and financed with ABCP (Acharya, Schnabl, and
Suarez 2012).
6.3.2
US Interest Rates
We have seen that the long- term upward trend in net foreign holdings of
US securities since 1994 has been the result of an upward trend in net foreign
holdings of US safe securities. The rise in net holdings by foreigners over
time has coincided with a downward trend in real interest rates, as illustrated
in figure 6.9. The real annual interest rate on the ten- year Treasury bond fell
from 3.78 percent at the start of 2000 to 1.97 percent by the end of 2005,
while the ten- year Treasury Inflation Protected (TIPS) rate fell from 4.32
percent to 2.12 percent over this period. Real rates fell further to all- time
lows during the housing bust. The real ten- year Treasury bond rate declined
from 2.22 percent to −0.42 percent from 2006:Q1 to 2011:Q3, while the TIPS
rate declined from 2.20 percent to 0.08 percent.
Fig. 6.9
US real interest rates
Sources: US Treasury, Survey of Professional Forecasters. From 2003 to 2011, US Treasury;
from 1997 to 2002, data are obtained from J. Huston McCulloch, http://www.econ.ohio- state
.edu/jhm/ts/ts.html. The complete sample runs from 1991:Q4 to 2011:Q3.
Notes: “10yr TIPS” is the yield on Treasury inflation protected securities (TIPS) adjusted to
constant maturities, corresponding to the third month of each quarter. “10yr CMT real” is
the ten- year constant maturity Treasury bond rate minus expectations of the average annual
rate of CPI inflation over the next ten years from the Survey of Professional Forecasters, in
percent per annum.
International Capital Flows and House Prices
261
Empirically, Bernanke, Reinhart, and Sack (2004) find evidence for lower
Treasury yields around periods of Japanese purchases of US Treasuries in
the 2000 to 2004 period, while Warnock and Warnock (2009) estimate that
twelve- month flows equal to 1 percent of GDP are associated with a 19
basis point reduction in long rates. They also find US mortgage rates to be
affected. The effects are large. Had the twelve months ending in May 2005
seen zero foreign official purchases of US Treasury and agency bonds, their
results suggest that, ceteris paribus, US long rates would have been about
80 basis points higher.
6.3.3
Financial Market Liberalization
While there is little doubt that inflows of foreign capital into US Treasury and Agency markets are associated with lower long- term real interest
rates, there is no direct evidence that they have played an important role in
raising house prices during the boom. We argued earlier that there are good
theoretical reasons to doubt the hypothesis that lower interest rates had a
quantitatively large effect on house prices during the boom. Empirically,
Glaeser, Gottlieb, and Gyourko (2010) concur and find that even when the
house price impact of lower interest rates should be stronger (at a low initial
rate), they account for, at most, one-fifth of the observed change in housing
prices. We present additional evidence on this in the following.
What, then, accounts for the dramatic rise in US house prices during
the boom if not low interest rates? A key missing element in this scenario
is the shift in credit standards and housing transactions costs, summarized
earlier as an FML and its reversal. The widespread relaxation of credit
standards is well- documented (see Nichols, Pennington-Cross, and Yezer
2005; Favilukis, Ludvigson, and Van Nieuwerburgh 2010; and Ferrero 2011
for more details). Moreover, a growing body of empirical evidence directly
links measures that identify changes in credit supply (as opposed to changes
in demand) to movements in asset prices.
Loan- to-Value Ratios
Many different aspects of mortgage lending over the 2000 to 2010 period
are consistent with a relaxation of credit standards. It may seem that an obvious way to measure relaxation of credit standards is to study loan- to-value
ratios. Several studies have observed that average or median loan- to-value
ratios did not increase much over time; see, for example, the contribution by
Glaeser, Gottlieb, and Gyourko in this volume (chapter 7). There are at least
three problems with using average LTV ratios as an indicator of tightness
of credit constraints. First, average loan- to-value ratio measures usually
mix in mortgages for house purchases with those for refinancing. The latter
category of mortgages have much lower LTV ratios because the borrowers often have accumulated substantial amounts of home equity already.
These refinancing are quantitatively important because, during the housing
boom, mortgage interest rates came down persistently, leading to a massive
262
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
refinancing boom. The share of refis in originations was 63 percent in 2002,
72 percent in 2003, and around 50 percent in 2004 to 2006 (data from www
.insidemortgagefinance.com).
Second, the average loan- to-value ratio are typically based only on the
first lien on the house. But often, new borrowers would take out an 80 percent LTV first lien and then a second (and possibly third) lien (closed- end
second or home equity line of credit). By the end of 2006 households routinely were able to buy homes with 100 percent or higher financing using a
piggyback second mortgage or home equity loan. The fraction of households with second liens rose dramatically during the boom. For subprime
loans, that fraction rose from 3 percent in 2002 to 30 percent; for Alt-A
loans it rose from 3 percent to 44 percent.12 In addition, second or third liens
were often the way in which existing home owners tapped into their home
equity, often several quarters after they took out the original mortgage.
This equity extraction through second liens is in addition to extraction via
cash- out refinancing, another innovation of the boom that became increasingly prevalent. The contribution in this volume by Lee, Mayer, and Tracy
(chapter 5) shows that second lien balances grew from about $200 billion
at the start of 2002 to over $1 trillion by the end of 2007. It also shows that
the prevalence of second mortgages rose in every US region from below 10
percent at the start of the boom (bit higher in coastal cyclical markets) to
around 40 percent in 2006 (except for the Midwest declining region, which
peaks at a 20 percent share).
What this evidence suggests is that we should look at combined LTVs
(CLTVs), combining all liens on a property, at the time of purchase. And to
gauge how credit constraints affected the marginal household, we should
look at the right tail of that CLTV distribution. Lee, Mayer, and Tracy (chapter 5, this volume) show that the average LTV at purchase for properties with
one lien stayed rather constant over the boom—if anything, it declined a bit.
Likewise, the share of purchases with one mortgage with an LTV greater or
equal to 95 percent also stays constant. By contrast, the share of purchases
with multiple mortgages with a CLTV greater or equal to 95 percent rises dramatically in every region. The nationwide increase is from about a 25 percent
12. An indirect indicator of the prevalence of the use of second mortgages is the fraction
of first liens with LTV exactly equal to 80 percent. This fraction rose substantially between
2002 and 2006, as shown by Krainer, LeRoy, and Munpyung (2009). They also show that the
fraction of FRMs with LTV greater than 80 percent decreased from 22 percent to 6 percent
over this period. Their hypothesis is that mortgage lending underwent a shift from a practice of
achieving greater home- buyer leverage by simply increasing the LTV on the first lien (common
prior to the housing boom), to a practice of achieving such greater leverage by combining an
exactly 80 percent LTV first lien with a second lien taken out simultaneously (common during
the housing boom). In short, during the housing boom high LTV ratios were achieved by taking
out piggyback second mortgages rather than by loading all leverage onto the first lien, as was
previous practice. Consistent with this hypothesis, Krainer, LeRoy, and Munpyung (2009) find
that the default rate on first lien mortgages with exactly 80 percent LTV ratios was higher than
that on first lien mortgages that had either 79 percent or 81 percent LTV ratios.
International Capital Flows and House Prices
263
Fig. 6.10 Fraction of properties in LA County with cumulative LTV ratios greater
or equal than 100 percent
Source: Laufer (2011).
Note: The figure plots the cummulative loan- to-value ratio at the time of purchase for homes
in Los Angeles County.
share to about a 60 percent share. At the peak, about two- thirds of purchase
mortgages with a second lien had a CLTV of 95 percent or more. Keys et al.,
chapter 4 in this volume, show that the average CLTV on subprime loans
increases from 80 percent in 1997 to 96 percent in 2006.
A recent study using detailed data on mortgages in Los Angeles county
shows the dramatic easing of credit constraints over the boom period and
subsequent reversal in another way. Figure 6.10 from Laufer (2011) shows the
share of properties in LA county with CLTVs at purchase above 100 percent
for all loans except nonconventional loans (Federal Housing Administration
[FHA] and Veterans Affairs [VA] loans). That share rises from 8 percent in
2001 to 54 percent in the fourth quarter of 2006, before collapsing. The sharp
drop in this series beginning in 2007 and reaching zero by 2008 reflects a
significant reduction in the maximum LTV ratio permitted by mortgage originators, since home values (in the denominator) were simultaneously falling.
Finally, there is a widespread belief that house price appraisals, done at
the time of mortgage origination, were upward biased during the boom. This
would downward bias LTV and CLTV ratios. As a result, what may look
like flat or modestly increasing (average) CLTVs may in fact be increasing
CLTVs once measured relative to the true value of the property.
Other Aspects of Credit Availability
The behavior of CLTV ratios in the boom and bust does not do full justice to several aspects of the increased availability of mortgage credit. New
264
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
mortgage products became available to borrowers that were previously unable
to obtain mortgage credit. The share of subprime mortgage originations (to
borrowers with low FICO credit scores) went from less than 10 percent of
originations in 2002 to 40 percent of originations by 2006, growing from
$120 billion in originations in 2001 to $600 billion in 2006 (www.insidemort
gagefinance.com). Likewise, the fraction of mortgages made to households
with debt- to-income ratios above 40 percent rose from 33 percent to 50 percent over the same period (Haughwout et al. 2011). The Alt-A market, which
grew from $60 billion in originations in 2002 to $400 billion in 2006, predominantly served households with low or no documentation (asset and income
verification). The fraction of Alt-A loans with full documentation declined
from 41 percent in 2002 to 19 percent in 2006. Complex mortgages, defined
by Amromin et al. (2011) as mortgages with low initial payments, grew from
about 2 percent of originations in 2002 to 30 percent of total originations
in 2006. Complex mortgages are non- fully amortizing loans, including the
interest- only mortgages studied by Barlevy and Fisher (2011), option ARMs
(pick- a- payment mortgages), negative amortization loans, loans with teaser
rates, and loans with balloon payments. Complex mortgages often went to
households with higher than average incomes, living in higher than average
expensive housing markets. In addition to making house purchases available
to some households that would otherwise not have been able to own a home,
complex mortgages may also have allowed other households to buy a larger
house than what they otherwise would have been able to afford.
Finally, private label securitization played an important role in providing
the funding for all these new mortgages. Keys et al. (chapter 4, this volume)
show that the fraction of subprime loans that was securitized increased from
about 50 percent in 2001 to 90 percent in 2007, before collapsing to 0 in 2008.
The fraction of conforming loans that were securitized also increased from
70 percent to 90 percent during the boom, and has stabilized at that level.
Exogenous Changes in Credit Supply
Moreover, a growing body of empirical evidence directly links measures
that identify changes in credit supply (as opposed to changes in demand) to
movements in asset prices.
Adelino, Schoar, and Severino (2012) exploit exogenous variation in the
government- controlled conforming loan limit (CLL) as an instrument for
changing credit supply. The CLL determines the maximum size of a mortgage that can be purchased or securitized by Fannie Mae or Freddie Mac.
Because these loans were widely understood to have the implicit (and since
2008, explicit) backing of the US government, borrowers in the market for
loans that fall below the CLL have easier access to credit at less costly terms.
Changes in the CLL are set annually and depend on the previous year’s limit
plus the change in the median national house price. These movements are
clearly exogenous to individual mortgage transactions, local housing mar-
International Capital Flows and House Prices
265
kets, and the local economy. Using data on single- family house purchases in
ten metropolitan statistical areas between 1998 and 2006, Adelino, Schoar,
and Severino show that houses that became newly eligible for a conforming
loan just after an increase in the CLL saw significant price increases relative
to similar houses that were already below the limit before the CLL increase.
Rajan and Ramcharan (2011a) exploit interstate banking restrictions to
study the effect of credit supply on land prices in the early twentieth century
United States. Regulations at the time stipulated that banks could not lend
across state borders. They argue that the number of banks in this era proxied
for credit supply, with more banks indicating higher supply. They show that
the number of banks in a county positively predicts land prices independently of fundamentals likely to move credit demand for land (commodity
prices). They also find that the number of banks in neighboring in-state
counties affects land prices more than the number of banks in equidistant
counties out of state. Since banks were prohibited from lending across state
borders, it is difficult to form a coherent story for this latter fact that does
not involve credit supply.
In a similar spirit, Favara and Imbs (2011) identify movements of credit
supply in more recent data (since 1994) by studying bank branching restrictions. Even though interstate banking (i.e., cross- state ownership of banks)
was made legal after the passage of the Interstate Banking and Branching
Efficiency Act of 1994, US states retained the right to erect barriers to interstate branching. They study branching deregulations since 1994 and show
that they significantly affect the supply of mortgage credit. With deregulation, the number and volume of originated mortgage loans rise, while denial
rates fall, echoing evidence in Mian and Sufi (2009). This deregulation has
no effect on a placebo sample, formed of independent mortgage companies
that should not be affected by the regulatory changes. Deregulation leads
to greater supply of mortgage credit, which they find leads to significantly
higher house prices.
Our main measure of credit availability is based on quarterly bank lending
surveys for countries in the Euro Area and the United States. For the United
States, we use the Senior Loan Officer Opinion Survey on Bank Lending
Practices (SLOOS), collected by the Federal Reserve. An important aspect
of this survey is that it asks banks to explicitly distinguish between changes
in the supply of credit as distinct from the demand for credit, on bank loans
to businesses and households over the past three months. Thus in principle, answers to the appropriate questions are able to identify a movement
in supply separately from a movement in demand. We focus on questions
related to mortgage credit supply to households. The detailed information
is considered highly reliable because the surveys are carried out by central
banks, which are also bank regulators with access to a large amount of
information about a bank’s operations, including those reflected in loan
applications and balance sheet data.
266
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
Fig. 6.11
Net percentage of US banks reporting easier credit standards
Source: US Survey of Senior Loan Officers, 1990:Q2 to 2011:Q2.
Notes: The figure reports the percentage of banks that reported easing of credit standards
on mortgages, less the percentage that reported tightening of standards. A positive number
indicates that more banks reported easing than tightening. A negative number indicates the
opposite (more banks tightening than easing). Until 2006, surveys did not distinguish between
prime and subprime mortgages. Beginning in 2007:Q1, the figure shows the net percentage
easing for two series: prime mortgages only, and a weighted average of prime and subprime
mortgages, where weights are computed based on the shares of prime and subprime in total
mortgages, as described in the appendix.
Data for other countries are from bank lending surveys conducted by
national central banks, and the European Central Bank (ECB). The survey
questions are modeled after the US Survey of Senior Loan Officers. (See
the appendix for data sources.) We use these data in the following empirical
analysis.
For the US SLOOS survey, banks indicate easing, tightening, or no change
in lending standards compared to the previous three months. We use the net
percentage of banks that have eased their lending standards on mortgage
loans as a measure of credit supply. This is the difference between the percentage of banks reporting easing and the percentage of banks reporting
tightening, thus a positive figure indicates a net easing of lending standards,
considering all bank respondents.
Figure 6.11 reports the net percentage of banks easing over time. We
denote this variable CSt. According to this measure, there was a significant
easing of standards from 2002 to 2006, and a very sharp tightening afterwards. Notice that this measure does not weight banks by their relative
importance in the mortgage market, nor does it weight the responses by the
degree of tightening. Thus, it is not an indicator of the strength of credit
International Capital Flows and House Prices
267
easing or tightening, only of its breadth. Moreover, until 2007, the survey did
not distinguish between prime and subprime mortgages. The figure shows
clearly a broad tightening of credit standards beginning at the end of 2006.
A cursory investigation of the figure suggests that the easing of standards
in the boom was more modest. One must be careful in interpreting this
series, however. There is a long string of observations starting in 1998 and
continuing through 2006 that shows a net easing of credit standards. Recall
that the survey asks banks about how their standards have changed relative
to the pervious three months. Thus a series of observations indicating easier
credit conditions relative to previous quarters by a few important banks in
the mortgage space, once cumulated, could indicate a significant relaxation
of underwriting standards.
We can relate CS to the growth in mortgages outstanding. Before doing so,
figure 6.12 shows the share of mortgages outstanding by holder, over time.
The line labeled “GSE portfolio and pools” are Agency and GSE- backed
Fig. 6.12
Mortgage shares by holder over time
Source: Federal Reserve, flow of funds, table L.218. Sample 1990:Q1 to 2011:Q2.
Notes: “ABS” is home mortgages by issuers of asset- backed securities that are special purpose
vehicles (SPVs), entities established by contractual arrangement to hold assets and to issue
debt obligations backed by the assets. “US Chartered Commercial Banks” refers to mortgages
held by US- chartered commercial banks. “Savings Institutions” refers to state- chartered savings banks, federal savings banks, cooperative banks, and savings and loan associations. “GSE
portfolio and pools” refers to mortgages held as MBS assets in the portfolio of governmentsponsored enterprises plus mortgages held in Agency- and GSE- backed mortgage pools not
on GSE balance sheets prior to 2010:Q1.
268
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
mortgage pools, comprised only by conforming mortgage loans.13 The line
labeled “ABS” refers to issuers of asset- backed securities. Issuers of assetbacked securities are special purpose vehicles (SPVs) of the non-GSE banking system, entities established by contractual arrangement to hold assets
and to issue debt obligations backed by the assets but moved off the balance
sheet of the parent company. Note that the mortgages held in ABS were
comprised entirely of nonconforming loans, since the conforming loans were
all held in the GSE portfolio and pools. The figure shows a significant change
in the composition of loans from 2002 to 2007: a sharp rise in the share of
ABS, which mirrors a sharp fall in the share of GSE loans. This indicates a
shift in the composition of mortgage lending, away from conforming debt
and toward nonconforming debt, a trend that was subsequently reversed
after 2007.
Table 6.1 shows that the short- term trends in CS are related to these
very changes in the composition of lending over the boom- bust period. We
investigate the relation between the four- quarter moving average value of
the SLOOS net percentage indicator CS shown in figure 6.11, and year- overyear growth in mortgage credit outstanding, by holder. The table reports
results from a regression of the latter on the former. The first column shows
the relation over the full sample 1991:Q1 to 2010:Q4. This column shows
that CS is positively related to growth in ABS and negatively related to
growth in mortgages held in GSE pools. The last row shows the results from
a regression of the ratio of ABS to GSE pools. Variable CS is positively
related to growth in this ratio. Thus the percentage of banks reporting an
easing of credit standards is associated with a shift in the composition of
loans, toward nonconforming loans and away from conforming loans. The
subsequent columns show that this result is unique to the boom- bust period
2000 to 2010. In both the boom (2000 to 2006) and bust (2007 to 2010), CS
is positively related to the ratio ABS/GSE, but it is negatively related to this
ratio in the years prior to this boom- bust cycle (1991:Q1 to 1999:Q4).
In the empirical work following, we will focus on the quarterly loan survey
data on mortgage credit standards as a measure of credit supply. In thinking about these data, it is instructive to consider how they may relate to the
notion of credit availability in FLVNa. In that model, an FML involves both
a reduction in transactions costs associated with buying and selling the housing asset, and a change in collateralized borrowing constraints. Consider the
borrowing constraint component, which takes the form:
(1)
i
(1 − ) PtH it+1,
−Bt+1
∀a,t
13. Prior to 2010:Q1, only a small fraction of GSE mortgage pools were held in portfolio
at Fannie Mae and Freddie Mac; most were held off- balance sheet in special purpose vehicles
(SPVs). Beginning in 2010:Q1, almost all Fannie Mae and Freddie Mac mortgage pools were
consolidated on Fannie Mae’s and Freddie Mac’s balance sheets as a result of new accounting
standards issued by the Financial Accounting Standards Board (FASB), statements 166 and
167, pertaining to Securitizations and Special Purpose Entities. We have consolidated the two
into a single series, labeled “GSE portfolio and pools.”
International Capital Flows and House Prices
269
i
where B t+1
is the amount of bonds household i owns at the beginning of
period t + 1, Pt denotes the relative price of housing in units of the nonhousing consumption good (Pt is the time t price of a unit of housing of fixed
i
quality and quantity), and H t+1
is the housing stock owned by household
i at the beginning of period t + 1. A negative value for Bi indicates a borrowing position. This equation represents the collateral constraint in the
model, where 0 1. It says that households may borrow no more than
a fraction (1 − ) of the value of housing, implying that they must post collateral equal to a fraction of the value of the house. This constraint can
be thought of as a down- payment constraint for new home purchases, but it
also encompasses collateral requirements for home equity borrowing against
Table 6.1
Mortgage holder
Regression of mortgage growth by holder on credit standards
1991:Q1–
2010:Q4
2000:Q1–
2006:Q1
2000:Q1–
2010:Q4
1991:Q1–
1999:Q4
All
0.024
(3.73)***
[0.24]
0.000
(0.00)
[–0.04]
0.033
(4.86)***
[0.40]
–0.006
(–1.41)
[0.00]
ABS
0.097
(3.91)***
[0.20]
0.356
(2.00)
[0.24]
0.125
(4.65)***
[0.44]
–0.259
(–4.69)***
[0.38]
Banks
0.019
(3.82)***
[0.10]
–0.022
(–0.26)
[–0.03]
0.025
(4.25)***
[0.17]
0.014
(0.92)
[0.01]
Savings
0.088
(3.50)***
[0.39]
0.160
(1.95)
[0.29]
0.101
(3.72)***
[0.45]
0.070
(2.22)**
[0.19]
GSE
ABS/GSE
–0.013
(–2.37)**
[0.11]
0.110
(4.76)***
[0.26]
–0.146
(–3.30)***
[0.53]
0.50
(2.41)**
[0.34]
–0.014
(–2.26)**
[0.15]
0.140
(4.78)***
[0.48]
–0.036
(–3.60)***
[0.25]
–0.217
(–4.68)***
[0.37]
Notes: Regressions of year-to-year growth (from t – 4 to t) of variable on column (1) on a
four-quarter moving average of CS = (CSt + CSt–1 + CSt–2 + CSt–3)/4 and a constant. Variable
CSt has been standardized; a positive value for this variable means that banks reported eased
credit conditions relative to the previous quarter. The row labeled “All” refers to regressions
of total home mortgages outstanding on the moving average of CSt. “ABS” refers to home
mortgages growth owned by issuers of asset-backed securities. “Banks” refers to growth in
mortgages held by US-chartered commercial banks. “Savings” refers to growth in mortgages
held by savings institutions. “GSE” refers to growth in mortgages held as MBS assets in the
portfolio of government-sponsored enterprises plus mortgages held in Agency- and GSEbacked mortgage pools not in GSE balance sheets prior to 2010:Q1. “ABS/GSE” refers to the
growth in the ratio of ABS to GSE. Data are from Federal Reserve, flow of funds table L.218.
***Significant at the 1 percent level.
**Significant at the 5 percent level.
270
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
existing homes. The constraint gives the maximum combined LTV ratio for
first and second mortgages and home equity withdrawal.
The paper by FLVNa asks how a plausibly calibrated change in its value
(along with a calibrated change in housing transactions costs) changes the
equilibrium outcome. Thus, one way credit supply can change is via a change
in the fraction (1 − ) of the home’s value that must be held as collateral.
But, as just discussed, borrowing capacity will fluctuate endogenously with
i
the collateral value Pt Ht+1
even if that fraction remains unchanged. This
represents an endogenous change in borrowing capacity, driven by economic
shocks and accompanied by revisions in expectations about future economic
conditions. These factors get immediately reflected in house prices, which
affects borrowing capacity and the tightness of the constraints. We argue
that either of these represent a change in credit supply in the sense that they
are related to a borrower’s access to funds via his or her credit constraint.
Moreover, the two could be correlated (expectations of a decline in economic
activity could lead to an increase in ), as they are in the transition dynamics
studied by FLVNa and displayed in figure 6.2.14
6.4
International Evidence on House Price Fluctuations
We have seen in the abovementioned that the United States experienced
large capital inflows and commensurate current account deficits at the same
time that it experienced strong growth in house prices. The same is true for
countries such as Iceland, Ireland, and Spain. In fact, during the boom,
there is a positive cross- country correlation of current account deficits with
house price growth on the one hand (Ferrero 2011; Laibson and Mollerstrom 2010; Gete 2010) and with value added and the labor share of the
construction industry on the other hand (Gete 2010). Using data that ends
before the bust, Aizenman and Jinjarak (2009) provide a precise estimate of
the relationship between house prices and external imbalances: a 1 standard
deviation increase in lagged current account deficits is associated with a 10
percent appreciation of real estate prices.
Panel A of table 6.2 replicates the flavor of these results for the boom
period and extends the sample to a larger set of countries. It reports real
house price growth (deflated by the Consumer Price Index [CPI]), cumulative current account deficits, and cumulative residential investment over the
period 2000:Q1 to 2006:Q4, with the last two variables measured relative
to GDP in 2006:Q4. Countries such as Germany, Switzerland, China, and
14. The transition dynamics are an exploration of movements between stochastic states with
different values for (as well as the housing transactions costs). An important aspect of the
transition is that the exogenous changes in borrowing capacity are correlated with endogenous
changes in borrowing capacity, because the exogenous and unexpected decline in is calibrated
to coincide with an economic boom, which bolsters collateral values and endogenously relaxes
borrowing constraints.
Table 6.2
Cross-country house price statistics
Panel A 2000:Q1–2006:Q4
Real HP gr.
(% change)
Australia
Austria
Belgium
Canada
China
Czech Republic
Denmark
Estonia
Euro Area
Finland
France
Germany
Greece
Hungary
Iceland
Ireland
Israel
Italy
Korea
Luxembourg
Netherlands
New Zealand
Norway
Poland
Portugal
Russia
Slovenia
Spain
Sweden
Switzerland
United Kingdom
United States
Corr. CA def.
Corr. HP gr.
55
1
18
46
–1
20
64
387
32
37
85
–16
50
40
64
60
–16
35
25
71
28
73
46
–2
–6
157
46
87
61
12
78
64
0.23
1.00
CA def.
(cum.)/
GDP2006
(%)
24
–8
–17
–10
–22
19
–16
47
0.04
–35
–3
–17
39
39
57
8
–6
7
–12
–51
–31
30
–73
18
51
–39
8
28
–35
–75
13
30
1.00
0.23
Panel B 2006:Q4–2010:Q4
Res. inv.
(cum.)/
GDP2006
(%)
35
29
32
35
38
18
32
19
39
36
38
42
25
28
57
27
29
29
13
37
35
21
16
42
8
11
45
17
28
21
32
0.22
–0.25
Real HP gr.
(% change)
CA def.
(cum.)/
GDP2006
(%)
17
20
10
10
–6
4
–20
–47
–3
8
1
–3
–22
–27
–28
–40
34
–2
–4
–3
–7
–10
9
33
2
10
1
–16
15
13
–6
–36
–0.38
1.00
23
–17
–1
6
–50
14
–15
26
2.4
–15
7
–29
62
15
62
15
–14
12
–9
–38
–27
29
–68
29
51
–30
18
35
–36
–40
9
17
1.00
–0.38
Res. inv.
(cum.)/
GDP2006
(%)
27
19
26
30
63
16
22
21
28
22
23
23
14
21
26
22
21
20
14
26
23
16
11
18
12
10
30
15
15
14
13
–0.14
–0.09
Notes: House prices are deflated by CPI. The data are from different national sources (see appendix), mostly quarterly,
except for Germany (annual), Luxembourg (annual, and until 2009), Italy (semiannual), and Japan (semiannual). The
CPI is collected by EIU from national sources. For Slovenia, series begins in 2003:Q1; for Russia in 2001:Q1. The CA
deficit data are from IMF, and SAFE for China. The CA balances are accumulated and deflated by 2006 GDP (collected
by EIU), all in current US dollars. For Belgium, CA data 2000:Q1 to 2001:Q4 are from NBB, via OECD. Residential
investment is from Eurostat and national sources. Residential investment is accumulated and deflated by 2006 GDP, all
in current national currency. For France, Hungary, Poland, Switzerland, and Russia, residential investment data are
available only through 2009.
272
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
Austria accumulated large current account surpluses and exhibited slow
house price growth and modest residential investment, while countries such
as the United States, Spain, the United Kingdom, Portugal, Greece, Estonia,
New Zealand, and Australia attracted lots of external capital, exhibited
large rises in house prices, and experienced significant residential investment booms. In the boom period, there is a positive cross- country correlation between average house price changes and average current account
deficits equal to 23 percent. There is also a negative cross- country correlation between residential investment and house price growth: countries with
more residential investment experienced lower house price growth, consistent with the idea that residential investment drives up the expected housing
stock and drives down the expected future growth rate on the dividend to
housing (rent).
It is tempting to conclude that the excess savings of the first group of
countries found its way to the real estate industry in the second group of
countries and fueled the housing boom there. However, as argued earlier,
general equilibrium considerations suggest that large inflows into safe assets
need not lead to large house price booms because the effect of lower interest
rates is offset by a rise in risk premia and an expected increase in the housing stock from higher residential investment. This may help explain why the
current account patterns from the boom period persisted in many countries
during the housing bust, while house prices and residential investment patterns obviously did not. Panel B of table 6.2 shows that the cross- country
correlation between house price changes and current account dynamics
reverses in the bust sample from 2006:Q4 to 2010:Q4. The cross- country correlation between the current account deficit and house price growth is now
−38 percent. By itself, this negative correlation is certainly not consistent
with the notion that capital inflows cause higher house price growth. Nor
is it consistent with the hypothesis that capital inflows lead to a relaxation
of credit standards, which in turn causes higher house price growth. To further explore these issues, we now turn to a statistical analysis of the relation
between house price changes (or changes in price- rent ratios), measures of
capital flows, credit standards, and interest rates.
6.4.1
Regression Analysis
In this section we undertake a basic empirical analysis of correlations
among house prices and other variables.
A few words about the data are in order. First, with regard to an international panel of data, we are limited to far fewer time- series observations
given the availability of bank lending survey data for non-US countries.
These data extend only from 2002:Q4 to 2010:Q4. For this reason we also
look separately at regressions for the United States alone, where data reaches
back much further, starting in 1990:Q4.
Second, as explained in the appendix, the data on bank lending standards
International Capital Flows and House Prices
273
differs somewhat by country. When we analyze the United States alone, we
use the net percentage easing indicator plotted earlier as a measure of credit
standards. For the other countries, the surveys are generally modeled after
the US SLOOS survey, but the way the survey results are aggregated can
differ. For nine of the eleven countries for which there exist data on credit
standards (Austria, Belgium, Euro Area, France, Korea, Portugal, Spain,
United States, Ireland) we have available, or can construct, a diffusion index
of credit standards or a scale transformation thereof (some countries report
a mean value indicator—see the appendix—which is a scale transformation
of the diffusion index) with the information reported by the central banks.
This diffusion index, however, is not a scale transformation of the net percentage indicator discussed earlier, and there are two countries (Canada and
the Netherlands) that report only the net percentage indicator. For the panel
regressions that we report on here, we simply use all these data together in
one regression, even though the credit standards measure for two countries
(Canada and the Netherlands) are not a scale transformation of the other
countries’ measures. Results, available on request, show that the findings are
virtually unchanged if we exclude these two countries. Finally, we standardize these bank lending survey measures, country by country, by subtracting
the mean and dividing by the standard deviation, for the period 2002:Q4
to 2010:Q4. This ensures that (at least for the nine countries for which we
have a diffusion index or mean value indicator) the credit supply measure
for each country is in the same units. The appendix provides more details
on the credit supply data by country.
Third, we use a measure of the change in net foreign holdings of US
securities as our main measure of international capital flows into the United
States. (Although for completeness, we also report results using the CA
[current account] deficit.) Annual net foreign holdings estimates are compiled by the BEA in their international investment data, year- end positions, thereby providing annual observations. All the rest of our data are
quarterly, however. Instead of limiting ourselves to annual observations,
we form an estimate of the quarter- end net foreign liability position of the
United States in total securities, by employing a methodology to interpolate
between the year- end positions, taking into account the quarterly transactions data in these same securities. The procedure ensures that our estimate
of the holdings at the end of the fourth quarter of a given year is equal to
the recorded value from the annual holdings data. The details of this procedure are described in the appendix. We simply note here that it provides
a quarterly measure of the change in net foreign holdings of US assets. We
use this measure in the regressions that follow.
We now present results from a regression analysis using these and other
data. We emphasize that, in presenting these next results, we do not make
claims about causality. Later we will provide some additional discussion and
evidence on the question of causality.
274
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
Table 6.3
Quarterly panel regressions (2002:Q4–2010:Q4), eleven countries
Real House price growth on
Regression
Cons.
CAdef/GDP
1
0.005
(1.52)
0.004
(1.69)
0.005
(1.62)
0.005
(1.58)
0.005
(1.96)
–0.055
(–0.73)
2
3
4
5
–0.018
(–0.29)
–0.009
(–0.14)
CS
(CAdef/GDP)xCS
R2
0.01
0.005
(3.24)***
0.005
(3.26)***
0.004
(3.20)***
0.06
0.07
0.083
(5.34)***
0.060
(6.61)***
0.05
0.09
Notes: Panel data estimation with fixed effects. CAdef/GDP is current account deficit divided
by the country’s GDP. Variable CS is net percentage of banks reporting easing of credit; a
positive value indicates more banks eased than tightened credit conditions with respect to
previous quarter. The column labeled “Cons” gives the coefficient on the regression constant.
Credit conditions have been standardized country by country. Driscoll–Kraay corrected tstatistics in parentheses (lags = 3). Eleven countries included are Austria, Belgium, Canada,
Euro Area, France, Ireland, Korea, Netherlands, Portugal, Spain, and the United States.
There are 363 observations in total.
***Significant at the 1 percent level.
We begin with evidence from the panel of eleven countries mentioned
earlier. For these countries, we have quarterly observations from 2002:Q4 to
2010:Q4 on real house price growth, the current account deficit, and credit
standards. The variable CS in these regressions is a diffusion index measure
that, when increased, indicates an easing of credit standards. A detailed
description of all of these data, including data sources, is given in the appendix. Table 6.3 reports the results of a panel fixed- effects regression of real
house price growth on the current account, CS, and interactions of these
variables. The variable CAdef/GDP is the current account deficit, divided
by the country’s GDP.
Table 6.3 shows that CAdef/GDP bears a negative relation to contemporaneous real house price growth (though it is not statistically significant),
suggesting that, if anything, capital inflows are associated with a decline in
house prices, rather than a boom.15 By contrast, the credit standards measure CS is statistically significant and positive, implying that an increase in
CS (an easing of credit standards) leads to an increase in real house price
growth (row 2). Row 3 shows that CS remains the only significant determinant of house price growth when both variables are included, while rows
15. We use the Driscoll and Kraay (1998) consistent covariance- matrix estimates to produce
heteroskedasticity consistent standard errors that are robust to general forms of autocorrelation and cross- sectional (spatial) correlation between the residuals.
International Capital Flows and House Prices
275
4 and 5 document some interaction effects: countries and time periods in
which there was an increase in credit supply experienced a larger increase in
house price growth if they also ran current account deficits. But, controlling
for this, CAdef/GDP has a statistically insignificant marginal effect on house
price growth (row 4), while CS by itself has a strongly significant marginal
effect (row 5). The R2 statistics range from 6 to 9 percent whenever CSt is
included in the regression, either by itself or interactively with CAdef/GDP.
These results provide little support for the hypothesis that capital inflows
played an important role in driving the changes in house prices internationally over the recent boom- bust period.
To interpret the magnitudes of the coefficients on CS, recall that this
variable is standardized, so a one- unit increase in this measure implies a 1
standard deviation increase around its mean. The coefficient is 0.005, which
implies that a 1-standard deviation increase in CS leads to a 50 basis point
rise in quarterly real house price growth, roughly a 2 percent rise at an annual
rate. This increase represents about one- quarter of a 1 standard deviation
change in quarterly US real house price growth (2.0 percent).
To investigate a longer time frame, we now turn to an analysis of US time
series data. Table 6.4 presents results from regressions using the same variables as in table 6.3, but this time only for the United States. The US data
are quarterly and span the period 1990:Q2 to 2010:Q4. Row 1 shows that
CAdef/GDP has no effect, by itself, on real house price growth. This variable
explains 2 percent of the quarterly variation in real US house price growth.
By contrast, CS is strongly statistically significant, and by itself explains
Table 6.4
Quarterly regressions for United States (1990:Q2–2010:Q4)
Real house price growth on
Regression
1
2
3
Cons.
–0.006
(–1.35)
0.001
(0.27)
–0.011
(–2.68)***
CAdef/GDP
CS
0.207
(0.92)
0.365
(2.54)**
Adj. R2
0.02
0.016
(9.94)***
0.017
(10.32)***
0.53
0.62
Notes: See table 6.3. The column labeled “Cons.” reports coefficients on the constant in the
regression. CAdef is current account deficit, GDP is gross domestic product, both from
the US Department of Commerce, BEA. Variable CS is a measure of credit standards from
the SLOOS survey that gives the net percentage of banks that reported easier credit conditions. A positive value for this variable therefore indicates an easing of credit conditions, while
a negative value indicates a tightening. We standardize the credit standards variable by dividing by the standard deviation and subtracting its mean based on data for the full sample.
Newey and West (1987) corrected t-statistics in parentheses.
***Significant at the 1 percent level.
**Significant at the 5 percent level.
276
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
53 percent of the quarterly variation in house price growth. When we include
both CAdef/GDP and CS in the regression, the current account now has a
statistically significant and positive effect, and this adds to the regression
model’s ability to explain the data: the adjusted R2 rises by 9 percentage points
to 62 percent. But this happens because, over this sample, the CS and CAdef/
GDP are again negatively correlated rather than positively correlated (capital
inflows are associated with a tightening of credit rather than an easing). Since
credit supply is so strongly positively related to house price growth, removing its effects by including it in the regression along with CAdef/GDP allows
the regression to distinguish a modest positive role for the current account.
To interpret the magnitude of the coefficients on CS reported in table 6.4,
recall that we standardize this variable so a one- unit increase is equal to a
1 standard deviation increase. The coefficient estimate in row 2 implies that
a 1 standard deviation increase in CS leads to a 0.016 unit (160 basis point)
rise in quarterly US house price growth, roughly a 6.6 percent increase at an
annual rate. This increase represents about three- quarters of a 1 standard
deviation change in quarterly US real house price growth (2.15 percent).
For comparison, table 6.5 shows output from the same regressions but
restricted to the subsample that only includes the recent boom- bust period:
2000:Q1 to 2010:Q4. The results with respect to the relation between CAdef/
GDP and house price growth are little changed: in this subsample the variable explains none of the variation in the growth of residential real estate
prices in a univariate regression. But credit standards explains a much
larger fraction of the variation in house price growth in this sample: CS
now explains 66 percent of the quarterly variation in house price growth
(row 2). The coefficient is also larger, equal to 0.019 in row 2. A 1 standard
deviation increase in CS in this subsample leads to a 190 basis point rise in
quarterly real house price growth, roughly a 7.6 percent increase at an annual
rate. This quantitatively large effect represents 88 percent of a 1 standard
deviation change in quarterly US real house price growth. Moreover, unlike
the results for the full sample, even when included in the regression along
Table 6.5
Quarterly regressions for United States (2000:Q1–2010:Q4)
Real house price growth on
Regression
1
2
3
Cons.
CAdef/GDP
–0.018
(–0.96)
0.007
(1.76)
–0.003
(–0.17)
0.435
(1.02)
Note: See table 6.4.
***Significant at the 1 percent level.
0.214
(0.57)
CS
Adj. R2
0.01
0.019
(11.43)***
0.019
(11.90)***
0.66
0.66
International Capital Flows and House Prices
Table 6.6
277
Quarterly regressions for United States (1990:Q2–2010:Q4)
Real house price growth on
Regression
Cons.
NFLt
1
0.003
(0.76)
0.001
(0.27)
0.000
(0.06)
–0.142
(–1.46)
2
3
0.036
(0.89)
CS
Adj. R2
0.06
0.016
(9.94)***
0.016
(8.75)***
0.53
0.53
Notes: See table 6.4. Variable NFLt is change in total net foreign US liabilities in total securities (where quarter-end positions have been estimated as described in the chapter), divided
by trend GDP. The trend is measured using a Hodrick and Prescott (1997) filter.
***Significant at the 1 percent level.
with CS, CAdef/GDP is statistically unrelated to house price growth in all
of the regression specifications over this subperiod. To summarize, to the
modest extent that the current account bears any relation to US house price
growth, it does so only in samples prior to the recent housing boom- bust.
There is no relation between these variables in the recent boom- bust cycle.
Returning to the full sample, table 6.6 shows the same regressions when
we replace CAdef/GDP with our quarterly measure of the change in the
net foreign holdings of total securities (the change in the US net foreign liability position in securities), divided by trend GDP. We denote this variable
NFLt. The results indicate that this variable has no effect on US house
price growth, whether it is included in the regression by itself or jointly with
CSt. The modest effect we found from current account deficits on house
price growth in the long US sample disappears once we replace the current
account deficit by a better measure of capital inflows.
So far we have investigated only contemporaneous correlations between
house prices, capital flows, and credit standards. Table 6.7 shows results from
forecasting regressions of real house price growth, ln(P) in period t + H, on
variables known at time t. (Analogous results for the price- rent ratio rather
than price growth are nearly identical and are available upon request.) We
report results from long- horizon forecasts of house price growth on CSt by
itself (row 1); on CSt, NFLt, and the real ten- year T-bond rate, r t10 (row 2);
and on CSt, NFLt, r t10, and the growth in real GDP, GDPt (row 3). The
first column of results reports results from the H = 0 ahead forecasts (contemporaneous correlations), and the following four columns report results
from the one-, two-, three-, and four- quarter ahead forecasts of these house
price measures.
Table 6.7 shows that credit standards are strongly statistically significantly
related to the change in the log of real house prices at all future horizons.
Indeed, this variable explains 47 percent of the one-, two-, and three- quarter
ahead variation, and 41 percent of the four- quarter ahead variation. By
278
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
Regressions of ln(Pt+H) on CS, covariates (1991:Q4–2010:Q4)
Table 6.7
US data
Forecast horizon H
Row
Regressors
Contemp.
1
2
3
4
1
CSt
2
CSt
0.015
(9.63)***
[0.52]
0.018
(6.29)***
0.036
(0.79)
–0.004
(–1.10)
[0.53]
0.017
(6.19)***
0.058
(1.10)
–0.005
(–1.23)
0.568
(1.60)
[0.54]
0.015
(7.00)***
[0.47]
0.016
(4.11)***
–0.026
(–0.40)
–0.003
(–0.72)
[0.48]
0.016
(4.30)***
–0.024
(–0.36)
–0.003
(–0.63)
0.036
(0.09)
[0.47]
0.028
(5.46)***
[0.47]
0.032
(3.87)***
0.018
(0.12)
–0.009
(–1.24)
[0.49]
0.033
(4.25)***
0.012
(0.07)
–0.008
(–0.99)
–0.153
(–0.19)
[0.48]
0.041
(4.76)***
[0.47]
0.054
(4.71)***
0.218
(1.04)
–0.019
(–2.33)**
[0.53]
0.054
(5.00)***
0.216
(0.87)
–0.019
(–2.00)**
–0.032
(–0.03)
[0.53]
0.050
(4.09)***
[0.41]
0.071
(4.57)***
0.435
(1.62)
–0.027
(–2.31)**
[0.50]
0.070
(4.67)***
0.456
(1.37)
–0.028
(–2.20)**
0.449
(0.27)
[0.49]
NFLt
r10t
3
CSt
NFLt
r10t
GDPt
Notes: Variable P is Core Logic National House Price Index. The column labeled “Contemp.” reports
coefficients from a regression of contemporaneous house price growth on variables; that is, ln(Pt) –
ln(Pt–1) on CSt. For all other columns, results are reported for forecasting regressions; for example,
ln(Pt+H) = ln(Pt+H) – ln(Pt) on CSt. Variable CS is net percentage of banks reporting easing of credit; a
positive value indicates more banks eased than tightened credit conditions with respect to previous quarter. Credit conditions have been standardized. Variable NFL is change in total net foreign holdings of
securities, adjusted as described in the chapter, divided by trend GDP. Variable r10 is real ten-year bond
yield: ten-year constant maturity rate in last month of quarter minus ten-year ahead inflation forecast,
median response, from Survey of Professional Forecasters. Variable GDP is real GDP growth. Newey–
West corrected t-statistics appear in parentheses. Adjusted R2 in brackets.
***Significant at the 1 percent level.
**Significant at the 5 percent level.
contrast, the other explanatory variables add very little to the explanatory
power of the house price forecasting regression. Moreover, the change in US
net foreign liabilities, NFLt, has a negative effect on house price growth,
one quarter ahead. The real interest rate does have a statistically significant
negative effect on house price growth in the three- and four- quarter ahead
regressions. The coefficient on the net foreign liabilities indicator NFLt is
statistically insignificant at every forecast horizon.
The evidence just presented indicates that house price growth is correlated
with lags of CSt. Case and Shiller (1989) have pointed out that house price
growth is correlated with its own lags. Since CSt is strongly contemporaneously correlated with house price growth, and since house price growth is
279
International Capital Flows and House Prices
Table 6.8
Quarterly long-horizon regressions
Panel A
ln(Pt+H) – ln(Pt) on
Forecast horizon H
Row
Regressors
1
2
3
4
1
log(HPt)
0.86
(9.29)***
[0.70]
1.58
(6.41)***
[0.65]
2.35
(5.85)***
[0.69]
2.95
(4.98)***
[0.63]
Panel B
ln(Pt+H) – ln(Pt) on
Forecast horizon H
1
eCS
0.76
(5.01)***
[0.23]
1.33
(4.08)***
[0.19]
2.07
(4.59)***
[0.22]
2.71
(4.42)***
[0.22]
Notes: See table 6.7. Results for US data 1991:Q4 to 2010:Q4. Variable lnPt+H is log of real
house price index, measured by the Core Logic index and deflated by the CPI; eCS is the residual from a regression of log(HPt) on CSt; CS is credit supply. Positive credit supply means
banks eased credit conditions with respect to previous quarter.
***Significant at the 1 percent level.
correlated with its own lags, it stands to reason that there may be significant
information overlap in lagged values of CSt and in lagged house prices for
future house prices. Indeed this is what we find, as table 6.8 shows, suggesting that part of the reason house price growth is correlated with its own lags
is that house price growth is correlated with lags of credit standards, and
credit standards matter for house prices. Panel A of table 6.8 shows that the
one- quarter lagged value of house price growth explains 70 percent of house
price growth next period. But panel B of table 6.8 shows that a residual from
a regression of house price growth on contemporaneous credit standards
CSt only explains 23 percent of next period’s house price growth. For the
four- quarter horizon, the residual only explains 22 percent while the raw
series explains 63 percent. This evidence suggests that the effects of credit
standards on house prices explain most (but not all) of the serial correlation
in quarterly house price growth.
To summarize, the evidence just discussed suggests that bank loan officers’
accounts of their willingness to supply more mortgage credit are strongly
statistically related to house price movements, both contemporaneously and
in the future, and both in the United States and in international data. By
contrast, data on capital flows, real interest rates, and GDP growth at best
add modestly to explanatory power of these statistical models, and most
of the time they are found to add nothing. We now turn to a discussion of
whether these movements in credit supply are exogenous to other factors.
280
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
6.4.2
Are Movements in Credit Supply Exogenous?
While the previous evidence strongly suggests that bank credit standards
and credit availability matter for home values, with increases in credit supply
associated with higher home prices, there is nothing in the evidence to suggest that such movements in credit standards are exogenous to the state of
the economy or to expectations about future economic conditions, including
the direction of future home prices. Nor is there any theoretical reason to
expect them to be. As in classic financial accelerator models, endogenous
shifts in collateralized borrowing capacity imply that economic shocks have
a much larger effect on asset prices than they would in frictionless environments without collateralized financing restrictions. Bank loan surveys on
credit standards could in principle elicit information on either or both forms
of a borrower’s access to funds, and we have no way of knowing from the
survey questions how much of any given change in standards is represented
by one or the other. In our view, either of these represents a movement in
credit availability, and both could be important for home prices.
Still, don’t endogenous movements in credit supply raise a question of
causality? If credit standards move in response to changing economic conditions, which in turn alter expectations of future home price movements,
then how do we rule out the possibility that (current and future) home prices
affect credit availability rather than the other way around? The answer, we
argue, is that we do not rule this out nor should we seek to, since the direction
of causality is not central for the question of whether credit availability plays
a role in driving asset price fluctuations. A natural benchmark is a complete
markets environment, where borrowing constraints and transactions costs
play no role in the equilibrium allocations. Indeed, it is hard to understand
why credit standards would be correlated at all with asset values in such an
environment. With incomplete markets, credit availability can have a large
dynamic impact on asset prices even if fluctuations in that availability are
completely endogenous. From this perspective, as long as we have a clean
measure of credit supply, as distinct from credit demand, any correlation
of credit supply with asset prices is evidence that credit supply matters for
asset prices.
These considerations lead to two questions. The SLOOS survey explicitly
asks banks to distinguish movements in credit supply from movements in
credit demand, but there may be some residual correlation between the two.
First, are movements in credit supply still associated with home price movements once we eliminate the correlation between credit supply and credit
demand? Second, does the credit supply measure that we study have an
exogenous component that still affects house prices once we control for
expectations about future economic conditions and once we take out possible linkages between international capital flows and credit supply?
Table 6.9 presents evidence pertaining to these questions. To do so, we first
Regressions of ln(Pt+H) on CS, covariates (1991:Q4–2010:Q4)
Table 6.9
US Data
Forecast horizon H
Row
Regressors
Contemp.
1
2
3
4
1
CSt
2
εCD,t
3
εCD,t
0.015
(9.63)***
[0.52]
0.015
(7.20)***
[0.48]
0.018
(5.28)***
0.023
(0.51)
–0.004
(–1.13)
[0.50]
0.013
(2.95)***
0.017
(0.38)
–0.002
(–0.51)
0.011
(2.32)**
[0.55]
0.014
(6.04)***
[0.39]
0.015
(4.98)***
[0.47]
0.015
(7.00)***
[0.47]
0.015
(5.50)***
[0.43]
0.015
(3.50)***
–0.038
(–0.55)
–0.003
(–0.76)
[0.44]
0.012
(2.11)**
–0.044
(–0.61)
–0.002
(–0.36)
0.008
(1.33)
[0.46]
0.014
(6.22)***
[0.41]
0.014
(3.90)***
[0.41]
0.028
(5.46)***
[0.47]
0.027
(4.39)***
[0.41]
0.031
(3.22)***
–0.012
(–0.08)
–0.009
(–1.22)
[0.43]
0.028
(2.30)**
–0.019
(–0.12)
–0.007
(–0.95)
0.008
(0.69)
[0.43]
0.029
(6.11)***
[0.48]
0.026
(3.25)***
[0.39]
0.041
(4.76)***
[0.47]
0.038
(3.94)***
[0.40]
0.052
(3.67)***
0.164
(0.71)
–0.020
(–2.10)**
[0.46]
0.051
(3.01)***
0.162
(0.70)
–0.019
(–1.99)
0.002
(0.15)
[0.45]
0.046
(7.85)***
[0.58]
0.038
(3.05)***
[0.39]
0.050
(4.09)***
[0.41]
0.047
(3.54)***
[0.35]
0.068
(3.58)***
0.368
(1.23)
–0.028
(–2.08)**
[0.43]
0.070
(3.10)***
0.372
(1.24)
–0.028
(–2.11)**
–0.003
(–0.15)
[0.42]
0.060
(8.15)***
[0.57]
0.048
(2.89)***
[0.36]
NFLt
r10t
4
εCD,t
NFLt
r10t
GDPt→t+4
5
εNFL,t
6
εGFL,t
Notes: See table 6.7. Variable lnPt+H is log of real house price index, measured by the Core Logic index
and deflated by the CPI; eCD,t is the residual from regressing CS (net percentage of banks’ reporting easing of mortgage credit) on a constant and the variable CD (net percentage of banks reporting higher
demand for mortgage credit). Variable eNFL,t is the residual from regressing CS on a constant, the variable
NFL, and 3 lags of NFL; eGFL,t is the residual from regressing CS on a constant, the variable GFL,
and 3 lags of GFL. The residuals have been standardized. Variable NFL is change in total net foreign
holdings of securities, adjusted as described in the chapter, divided by trend GDP. Variable GFL is
change in total gross foreign holdings of securities, adjusted in the same manner as NFL, divided by
trend GDP; r10 is real ten-year bond yield: ten-year constant maturity rate in last month of quarter minus
ten-year ahead inflation forecast, median response, from Survey of Professional Forecasters. Variable
GDPt→t+4 is median forecasted real GDP growth between periods t and t + 4, from the Survey of Professional Forecasters. Row 1 is repeated from table 6.7 for ease of comparison. Newey–West corrected tstatistic in parentheses (lags = 4). Adjusted R2 in brackets. Seventy-seven quarterly observations.
***Significant at the 1 percent level.
**Significant at the 5 percent level.
282
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
regress the raw credit supply series CSt on the SLOOS survey’s measure of
credit demand (the net percentage of banks reporting higher credit demand),
and take the residual of this regression εCD,t, as a measure of credit supply.
(After obtaining this residual we standardize it so as to give it the same
units the raw measure used previously had.) We also replace current GDP
growth from the previous regressions with the expected GDP growth rate
for the year ahead, GDPt→t+4, as measured by the Survey of Professional
Forecasters (SPF) median forecast. In these regressions, we continue to use
our measure of capital inflows, NFLt, and the real ten- year T-bond rate,
10
r 10
t , as additional explanatory variables. Notice that r t is itself a forwardlooking variable since it equals the nominal ten- year T-bond rate minus the
expected ten- year inflation rate (also from the SPF, median forecast). Thus,
once we include these forecasts of future economic activity and r 10
t as additional predictor variables, any remaining role for our residual credit supply
measure εCD,t in explaining house price movements must be independent of
expectations of future real activity or inflation.
Table 6.9 shows that the residual credit supply measure εCD,t by itself
explains about the same amount of variation in house price growth as does
the raw series CSt (given in row 1), and this is true no matter what other variables we include as additional regressors. The table also shows that expected
future GDP growth GDPt→t+4 has significant explanatory power for house
price growth contemporaneously but does not help predict future house
price growth, consistent with the notion that such expectations are reflected
immediately in asset prices and collateral values. Even so, the residual supply
measure εCD,t maintains its marginal explanatory power for contemporaneous movements in ln(P). When it comes to forecasting future house price
changes, only the residual credit supply measure εCD,t displays any clear predictive power: expectations of future GDP growth, real interest rates, and
the change in US net foreign liabilities all have a statistically negligible effect
of ln(Pt). The change in US net foreign liabilities, NFLt, again has a negative (but statistically insignificant) effect on house price growth, one and two
quarters ahead. The forecasting regressions for horizons ranging from one to
four quarters ahead using all four predictor variables explain about the same
amount of variation in future house price growth as does the residual credit
supply measure alone, indicating that none of them are strongly related to
future house price growth once we control for credit supply.
A related question on the exogeneity of credit supply movements concerns the relationship between international capital flows and credit supply.
Could international capital flows have been a causal factor contributing to
the changing supply of credit, thereby indirectly contributing to the housing
boom and bust through their influence on credit supply? The remaining rows
of table 6.9 provide evidence on this question. To address this question, we
again construct residual credit supply measures by regressing the raw credit
supply CSt series on the contemporaneous values and three quarterly lags
International Capital Flows and House Prices
283
of two different measures of international capital flows: NFLt, and an
analogous measure of the change in gross flows, GFLt, constructed in the
same way as NFLt except that we do not net out US holdings of foreign
assets from the foreign holdings data. We then take the residuals from these
regressions, denoted εNFL,t and εGFL,t respectively, as measures of changes in
credit supply that are unrelated to current and past changes in international
capital flows.
Rows 5 and 6 present the results from regressions of house price growth
on εNFL,t and εGFL,t. Comparing row 1 with rows 5 and 6, we see that these
residual credit supply measures explain almost the same amount of variation
in house price growth as does the raw series CSt, indicating that the bulk of
the credit supply variation that is related to house price movements is in fact
orthogonal to movements in capital flows. Moreover, the first- stage regression results (not reported) indicate that current and lagged credit flows have
a negative effect on credit supply CSt, contradicting the hypothesis that credit
flows into the United States contributed to an easing of credit supply during
the housing boom, and vice versa during the housing bust. In short, there
is no evidence from these data to support the hypothesis that international
capital flows were causally related to the changes in credit supply that were
responsible for the housing boom and bust.
To summarize, when we control for expectations about future economic
conditions and when we purge the credit standards measure of any residual demand or capital flow effects, we still find that credit supply has an
economically large effect on house price movements. Indeed, these exogenous movements in credit supply appear to be the main driver of the credit
supply- house price link we find. We close this section by noting that the
model in FLVNa produces qualitatively similar results when performing
such regressions on simulated data. To conserve space, these results are not
reported but are available upon request.
6.5
Conclusion
In this chapter we have studied the empirical relationship between house
price changes and international capital flows, focusing in particular on the
extraordinary boom- bust period in the housing market from 2000 to 2010.
We have argued that foreign capital flows into safe US securities—US Treasury and Agency bonds—played an important role in understanding the
low interest rates in the last decade and quantitatively account for all of the
upward trend in the US net foreign liability position over this period. Many
countries that saw large housing booms and busts attracted foreign capital,
as witnessed by their large current account deficits. Much of this capital
seems to have found its way into residential investment and mortgage credit
extension.
Despite these stylized facts, we have argued here that the same capital
284
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
inflows that lowered interest rates and supported the mortgage boom over
this period had only a small impact on house prices. Although the housing boom was characterized by sharp increases in the rate at which capital
flowed into the United States, the bust and its aftermath occurred with no
clear reversal in the trend toward healthy capital inflows into US assets
considered to be safe stores- of-value and integral to housing finance. While
US borrowing from abroad may ultimately decline, home values have not
waited to do so, having already given up almost all of their gains during the
boom years. This simple observation is reflected in our statistical analysis
of the relationship between home prices, capital flows, credit standards,
and interest rates, not only in the United States, but also internationally:
capital flows have little if any explanatory power for residential real estate
fluctuations in samples that include both the boom and the bust. We have
also argued on theoretical grounds that capital inflows—even those that
significantly decrease domestic interest rates—need not have large effects
on domestic real estate prices if they simultaneously push up the housing
risk premium and the expected stock of future housing.
A quantitatively meaningful account of the massive boom and bust in
house prices must therefore rely on (at least one) alternative element. We
argue here that the missing element is the financial market liberalization
in the mortgage space (and its subsequent reversal), which made it easier
and cheaper during the boom period for homeowners to purchase a house
or to borrow against existing home equity. The relaxation of credit constraints, by itself, is a powerful force for higher house prices in general equilibrium theory. Easier access to mortgage credit increases households’ ability
to withstand income shocks, and it reduces the risk premium households
require to invest in risky assets like houses. In addition, lower transaction
costs associated with new or refinanced home mortgages and home equity
lines of credit raise the liquidity of houses, and therefore their price.
We have presented evidence that these mechanisms appear to have operated in the United States, but also in countries other than the United States.
Using observations on credit standards, capital flows, interest rates, and
house prices, we find that, of these variables, only the bank lending survey
measure of credit standards explains either current or future home price
fluctuations, with credit supply explaining 53 percent of the quarterly variation in US house price growth over the period 1992 to 2010, and 66 percent
over the boom- bust period from 2000 to 2010. By contrast, the other variables combined add less than 5 percent to the fraction of quarterly variation
in house price changes explained, once we control for credit standards.
Our chapter is silent on the origins of the financial market liberalization in credit standards and its subsequent reversal, but it is worthwhile to
conclude by briefly considering some possibilities. A first possibility is that
mortgage lenders were confronted with exogenous changes in technology
International Capital Flows and House Prices
285
that affected mortgage finance. The boom period was characterized by a
plethora of such changes, including the birth of private label securitization,
collateralized debt obligations, credit default swaps, and automated underwriting coupled with new credit scoring techniques employed in that underwriting (Poon 2009). These specific innovations have been directly linked
to the surge in mortgage credit and house price growth by Mian and Sufi
(2009) and Keys et al. (chapter 4, this volume). Second, during the period
leading up to and including the housing boom, numerous legislative actions
were undertaken that had the effect of giving banks far more leeway to relax
lending standards than they had previously. In this regard, Mian, Sufi, and
Trebbi (2010) mention 700 such housing- related legislative initiatives that
Congress voted on between 1993 and 2008, while Boz and Mendoza (2010)
highlight the 1999 Gramm-Leach-Bliley and the 2000 Commodity Futures
Modernization Acts. Third, in the period leading up to and including the
housing boom, regulatory oversight of investment banks and mortgage
lenders weakened substantially (Acharya and Richardson 2011). For example, the regulatory treatment of AA- or- better rated private label residential mortgage- backed securities was lowered in 2002 to the same lesser
regulatory capital level as that which applied since 1988 to MBS issued
by the Agencies. Overall, regulatory capital requirements were in practice
further relaxed as banks ostensibly removed MBS assets from their balance sheets by selling them to SPVs with far less restrictive capital requirements than those assets would have faced had they remained on the bank
balance sheets. The removal of these assets from bank balance sheets was
illusory, however, since banks extended credit and liquidity guarantees to
the SPVs, effectively undoing the risk- transfer (but not the lowered capital
requirement) and forcing them to reclaim the MBS on their balance sheets
when those assets began to lose value (Acharya, Schnabl, and Suarez 2012).
These changes took place in an environment where private sector mortgage
lenders where engaged in a “race to the bottom” in credit standards with
the government- sponsored enterprises, which experienced similar regulatory changes and increasing investor awareness of the implicit government
guarantees that were afforded these enterprises (Acharya, Richardson,
et al. 2011). Faced with such changes in their regulatory environment, we
hypothesize that prior to the boom, mortgage lenders formed expectations
of higher future house price growth, leading to more and riskier mortgages
in equilibrium, as in the optimal contracting framework of Piskorski and
Tchistyi (2010). The bust was characterized by a tightening of regulatory
oversight (Acharya, Cooley, et al. 2011), which appears to have contributed
to the reduction in credit availability that occurred at the end of the housing boom. The exact mechanisms through which credit standards were
dramatically relaxed and reversed remains an important topic for future
research.
286
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
There are several other interesting questions for future research. First, it
is often hypothesized that international capital flows themselves contributed to a relaxation of lending standards during the housing boom. As
we show here, however, this explanation is not consistent, especially in the
United States, with the housing bust period, in which credit standards dramatically tightened but capital inflows to US safe securities remained high
on average and real interest rates low. Neither net capital flows nor gross
capital flows into the United States have any explanatory power for current
or future bank lending standards in our sample. More research is needed
on this question. Second, why is capital flowing from relatively productive
economies, like China, Germany, Japan, Switzerland, and so forth, to relatively unproductive economies like Spain, the United States, Greece, and
Italy? Moreover, why is capital flowing into safe assets like US Treasuries?
We have argued here that purchases of US safe assets appear to be driven by
reserve currency motives and political constraints by governmental holders
in the source countries (see Alfaro, Kalemli-Ozcan, and Volosovych 2011),
but more research is needed on this issue as well. Finally, for the most part
we focused on the relationship between net capital flows and house prices.
Other researchers (notably Obstfeld 2011), have argued that gross flows in
international financial assets may lead to financial market instability. Future
work should investigate the link between gross flows and prices of all kinds
of assets, including real estate, equity, and bond markets.
Appendix
This appendix provides details on all the data used in this study, including
data sources. The last section also includes some additional details about
the estimation procedures used.
House Price Data
Data on house prices are deflated by a consumer price index (CPI). The
data are from different national sources, mostly quarterly, except for Germany (annual), Luxembourg (annual, and until 2009), Italy (semiannual),
and Japan (semiannual). The CPI is collected by Economist Intelligence
Unit (EIU) and from national sources. For Slovenia the series begins in
2003:Q1; for Russia in 2001:Q1.
United States
For regressions involving the house prices in the United States, we use
the Core Logic National House Price Index (SA, Jan. 2000 = 100). This is a
International Capital Flows and House Prices
287
repeat- sales price index that is based on the universe of mortgages (conforming and nonconforming).16 House prices are deflated using consumer price
index (All urban consumers, US city average, All items) from the Bureau of
Labor and Statistics (http://www.bls.gov/cpi/). The monthly data are averaged over the quarter and rebased at 2005 = 100, so real house prices are
in 2005 US dollars. Regressions of growth rates use log changes, log(HPt)
− log(HPt−1) for contemporaneous changes, and log(HPt+H ) − log(HPt) for
H horizon changes.
For regressions using the aggregate price- rent ratio, we construct an index
by combining a measure of rent, for primary residences, constructed from
the Shelter component of the Consumer Price Index for all urban consumers, SA, last month of each quarter, with the Core Logic measure of house
prices. Data for rent are available from the Bureau of Economic Analysis of
the US Department of Commerce. The price- rent ratio has been normalized
to equal the value in 1975:Q4 of the quarterly price- rent ratio constructed
from the flow of funds housing wealth and National Income and Products
data on housing consumption.
International Data
House prices are deflated using consumer price indices for each country,
from national sources. For the Euro Area we deflate with the Harmonized
Index of Consumer Prices for the Euro Area 17, all items. Data sources for
residential real estate are given in table 6A.1.
Current Account Data
Current Account is measured as the current account deficit. Data are
available from the International Monetary Fund (IMF) for all countries
except China. For China, data are from State Administration of Foreign
Exchange, SAFE. The CA balances are accumulated and deflated by 2006
GDP (collected by Economist Intelligence Unit, EIU), in current US dollars. For Belgium, CA data 2000:Q1 to 2001:Q4 are from the National Bank
of Belgium, provided by the Organization of Economic Cooperation and
Development (OECD).
United States: (CAdef/GDP)t is current account deficit over nominal
GDP, at current market prices. Balance of current account is from the
Bureau of Economic Analysis, US International Transactions Accounts
Data, in millions of dollars, not seasonally adjusted. The GDP data is from
the Bureau of Economic Analysis, National Economic Accounts, in billions
16. Other indexes are available only for conforming mortgages. For example, the Federal
Housing Finance Administration (FHFA) measure is based only on conforming mortgages
and therefore misses price changes associated with nonconforming mortgages. Like the Core
Logic measure, the Case–Shiller measure is also based on the universe of mortgages, but it has
substantially smaller geographic coverage than the Core Logic measure.
Table 6A.1
Australia
Austria
Belgium
Canada
China
Czech Republic
Denmark
Estonia
Euro Area
Finland
France
Germany
Greece
Hungary
Iceland
Ireland
Israel
Italy
Korea
Luxembourg
Netherlands
New Zealand
Norway
Poland
Portugal
Russia
Slovenia
Spain
Sweden
Switzerland
Data sources for house prices
Australian Bureau of Statistics. Exist. dwellings (8 CITIES), PER DWEL.,
Q-ALL NSA. (1)
Oesterreichische National Bank. All dwellings (VIENNA), PER SQ.M., Q-ALL
NSA. (1)
EXISTING DWELLINGS, PER DWEL., Q-ALL NSA. (1)
Teranet-National Bank of Canada. Comp. 6 Cities (monthly, averaged to
quarterly using sales pair count).
National Bureau of Statistics of China. Land prices, Resid. and Commercial,
Q-ALL NSA. (1)
Czech Statistical Office. House prices.
Statistics Denmark. ALL SINGLE-FAMILY HOUSE, PER DWEL, Q-ALL
NSA.
Statistical Office of Estonia. Av. price per sq.m., 2-rooms and kitchen, Tallinn
(1). 2009 onwards, 55–70m2.
European Central Bank. Euro area 17 (fixed composition); New and existing
dwellings; Not S.A.
Statistics Finland. EXISTING HOUSES, PER SQ.M, Q-ALL NSA. (1)
National Institute of Statistics and Economic Studies. Existing Dwellings,
Q-ALL NSA. (1)
Deutsche Bundesbank, based on data provided by BulwienGesa AG. Existing
Dwellings, Y-ALL NSA. (1)
Bank of Greece. ALL DWELL. (URBAN GREECE EX. ATHENS), PER
SQ.M, NSA. (1)
FHB Bank. FHB House Price Index (actual buying and selling transaction data
of residential real estate).
Icelandic Property Registry. ALL DWELLINGS (GR. REYKJAVK), PER
SQ.M, M-ALL NSA. (1)
Economic and Social Research Institute (ESRI). ALL DWELLINGS, PER
DWELLING., Q-ALL NSA. (1)
Central Bureau of Statistics. Prices of Dwellings (Until September 2010–OwnerOccupied Dwellings).
Bank of Italy. ALL DWELLINGS, PER SQUARE M., H-ALL NSA. (1)
Kookmin Bank in Korea. ALL DWELLINGS, M-ALL NSA. (1)
Central Bank of Luxembourg. ALL DWELLINGS, Y-ALL NSA. (1)
Statistics Netherlands. Existing own homes. Dwellings: all. Price index purchase
prices.
Reserve Bank of New Zealand. All dwellings, Q-ALL NSA. QVL. (1)
Statistics Norway. All dwellings, Q-AVG, NSA. (1)
Central Statistics Office. Price of a square meter of usable floor space of a
residential building.
Inteligência de Imobiliário. All dwellings, PER SQUARE METER, M-ALL
NSA. (1)
Federal State Statistics Service. EXISTING DWELLINGS, PER SQUARE M,
Q-ALL NSA. (1)
Statistical Office of the Republic of Slovenia. Existing Dwellings, Q-ALL NSA.
(1)
Bank of Spain. All dwellings, PER SQUARE M., Q-ALL NSA. (1)
Statistics Sweden. ALL OWNER-OCCUPIED DWELLINGS, PER DWEL.,
Q-ALL NSA. (1)
Swiss National Bank. ALL 1-FAMILY HOUSES, PER DWELLING, Q-ALL
NSA. (1)
International Capital Flows and House Prices
Table 6A.1
United Kingdom
United States
289
continued
Office for National Statistics. All dwellings (ONS), PER DWEL., M, Q-ALL
NSA. (1)
Federal Housing Finance Agency. Family Houses, Q-ALL NSA (all transactions,
FHFA). (1)
Notes: Series (1) can be found at the Bank of International Settlements website, http://www.bis.org
/statistics/pp.htm.
of current dollars, seasonally adjusted at annual rates; we transform it to
quarterly rates dividing by four.
International data: (CAdef/GDP)t is current account deficit over nominal
GDP, at current market prices. Data from International Monetary Fund,
in millions of US dollars, and GDP data are collected by the Economist
Intelligence Unit on nominal GDP (USD), quarterly.
Residential Investment
Residential Investment data are from Eurostat and national sources, as
indicated in table 6A.2. Residential investment is accumulated and deflated
by 2006 GDP, all in current national currency. For France, Hungary, Poland,
Switzerland, and Russia, data is available only until 2009.
Data on Credit Standards
United States
The variable CSt is a net percentage index that indicates the percentage of
banks relaxing credit standards for mortgage loans (both considerably and
somewhat), with respect to the previous quarter, minus the percentage of
banks tightening credit standards (both considerably and somewhat). This
indicator is taken from the Senior Loan Officer Opinion Survey on Bank
Lending Practices for the United States, published by the Federal Reserve.
They report the net percentage of banks tightening standards. The negative of this is the net percentage of banks easing standards, which we use
in our empirical work. This is a quarterly survey of approximately sixty
large domestic banks and twenty- four US branches and agencies of foreign
banks. Questions cover changes in the standards and terms of the banks’
lending and the state of business and household demand for loans. These
data are available since May 1990, when the survey then began including
approximately twenty questions designed to measure changes in credit standards and terms on bank loans and perceived changes in the demand for
bank credit. See http://www.federalreserve.gov/boarddocs/snloansurvey/.
Table 6A.2
Australia
Austria
Belgium
Canada
China
Czech Republic
Denmark
Estonia
Euro Area
Finland
France
Germany
Greece
Hungary
Iceland
Ireland
Israel
Italy
Japan
Korea
Luxembourg
Netherlands
New Zealand
Norway
Poland
Portugal
Russia
Slovenia
Spain
Sweden
Switzerland
United Kingdom
United States
Data sources for residential investment
Australian Bureau of Statistics. Private Gross fixed capital
formation, Dwellings, Total, $ millions
Eurostat.
National Bank of Belgium. Gross fixed capital formation, dwellings,
current prices, millions of euros.
Statistics Canada. Capital formation in residential structures, current
prices, national currency, NSA.
National Bureau of Statistics China (NBSC). Total Inv. Residential
Buildings in the whole country, million yuan.
Eurostat.
Eurostat.
Eurostat.
No data.
Eurostat.
Eurostat.
Eurostat.
Eurostat.
Eurostat.
Eurostat.
Central Statistics Office Ireland. Gross Dom. Physical Cap.
Formation at current prices. Fixed capital, Dwellings.
Central Bureau of Statistics. Gross Domestic Capital Formation.
Residential Buildings, current prices.
Eurostat.
Cabinet Office, Gov. of Japan, Billions of Yen, nominal Private
residential investment, not SA.
Bank of Korea, Gross fixed capital formation in residential
buildings, current prices, Bil. Won.
Eurostat.
Eurostat.
Statistics New Zealand. Gross Fixed Capital Formation, Residential
buildings, current prices, $ millions.
Norway Statistics. Gross fixed capital formation, Dwellings
(households). Current prices (mill. NOK)
Eurostat.
Eurostat.
Federal State Statistics Service. Investment in fixed capital, in
residential houses, current prices, billions rubles.
Eurostat.
Eurostat.
Eurostat.
Eurostat.
Eurostat.
US Bureau of Economic Analysis. Fixed Investment, Residential,
Billions of dollars.
Notes: Most of the series are from national sources via Eurostat. Gross fixed capital formation, in construction work: housing, millions of national currency, current prices.
International Capital Flows and House Prices
291
We focus on the question that asks about residential mortgage loans at each
bank. From 1990:Q2 (beginning of the survey) to 2006:Q4, the question is
about residential mortgage loans in general:
Over the past three months, how have your bank’s credit standards for
approving applications from individuals for mortgage loans to purchase
homes changed?
The recommendations for answering this question state
If your bank’s credit standards have not changed over the relevant period,
please report them as unchanged even if the standards are either restrictive or accommodative relative to longer- term norms. If your bank’s
credit standards have tightened or eased over the relevant period, please
so report them regardless of how they stand relative to longer- term
norms. Also, please report changes in enforcement of existing standards
as changes in standards. (See http://www.federalreserve.gov/boarddocs
/snloansurvey/ for more details.)
From 2007:Q1 onwards, the question is asked for each of three categories
of residential mortgage loans: prime residential mortgages, nontraditional
residential mortgages, and subprime residential mortgages. The answer to
this question can be one of the following: tightened considerably, tightened somewhat, remained basically unchanged, eased somewhat, eased
considerably. Responses are grouped in “Large Banks” and “Other Banks.”
The index, however, is calculated using information of “all respondents.”
Given that the question is referenced to the past three months, we date
the index with respect to the quarter when changes to lending standards
occurred (as opposed to when the responses are collected, i.e., net percentage reported in July 2011 is the net percentage for 2011:Q2). In the report
beginning in 2007:Q1, a distinction is made between prime and subprime
mortgages in the survey. In the regressions using US credit supply, CS is a
weighted average of prime and subprime mortgages: (net percentage easing
on prime)∗weight plus (net percentage easing on subprime)∗(1 − weight),
where weight is 0.75 for 2007 and 0.95 for 2008. After that weight equals
1, because no bank reported that they originated subprime mortgages. The
earlier weights are based on the paper http://www.jchs.harvard.edu/publi
cations/finance/UBB10-1.pdf, page 85, figures 1 through 3. These numbers
are approximately the average share of banks that originated subprime residential mortgages, according to the survey (23 percent for 2007 and 8 percent
for 2008). Results are not sensitive to using one or the other set of numbers.
We standardize the net percentage indicator by subtracting the mean and
dividing by the standard deviation.
292
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
International Data
Variable CSt stands for credit standards for housing loans. Data are from
bank lending surveys conducted by national central banks and the European Central Bank. The survey questions are modeled after the US Survey
of Senior Loan Officers. Central Banks report the information in different
ways. Some Central Banks report net percentages, some report diffusion
indices, and some report mean values. Net percentage is the percentage of
banks loosening (or tightening) credit standards with respect to the previous quarter minus the percentage of banks tightening them (or relaxing).
Diffusion index is the percentage of banks loosening (or tightening) credit
standards “considerably” with respect to the previous quarter multiplied by
1 plus the percentage of banks loosening (or tightening) credit standards
“somewhat” multiplied by 0.5 minus the percentage of banks tightening (or
relaxing) “somewhat” times 0.5, minus the percentage of banks tightening
(or relaxing) “considerably” times 1. Mean values: each answer receives a
value from 1 to 5 (where for 5, the bank reported that relaxed the credit
standards “considerably,” 3 did not change them, and 1 is a “considerable”
tightening), and the mean value for each quarter is reported.
Mean values are a scale transformation of the diffusion index, but the
net percentage indicator is not. There are nine countries for which we can
construct either the diffusion index (Austria, Belgium, Euro Area, France,
Korea, Portugal, Spain, United States) or a mean value (Ireland) with the
information reported by the central banks. We have a larger set of eleven
countries for which we have only information on the net percentage (Canada
and the Netherlands). We standardize these indices, country by country, by
subtracting the mean and dividing by the standard deviation, for the period
2002:Q4 to 2010:Q4.
A positive value for CSt reflects easing credit conditions with respect to
the previous quarter, and units are in terms of standard deviations.
Micro data for each country are not publicly available, but each of the
abovementioned countries publishes an indicator (net percentage, mean
index, diffusion index) that reflects the change in credit conditions in the
country. The data sources are summarized in table 6A.3. For Austria,
Belgium, Euro Area, France, Ireland, Netherlands, Portugal, and Spain,
the survey is based on the Bank Lending Survey conducted by the ECB.
See http://www.ecb.int/stats/money/surveys/lend/html/index.en.html. The
European Central Bank’s website states:
The survey addresses issues such as credit standards for approving loans
as well as credit terms and conditions applied to enterprises and households. It also asks for an assessment of the conditions affecting credit
demand. The survey is addressed to senior loan officers of a representative
sample of euro area banks and will be conducted four times a year. The
sample group participating in the survey comprises around 90 banks from
International Capital Flows and House Prices
293
all euro area countries and takes into account the characteristics of their
respective national banking structures.
We focus on question 8 from the ECB survey, Item 8.1, Loans for house
purchase:
Over the past three months, how have your bank’s credit standards as
applied to the approval of loans to households changed?
Respondents can reply with one of the following answers: tighten considerably, tighten somewhat, basically unchanged, ease somewhat, or ease
considerably.
For Korea and Canada, the raw questions in the survey differ somewhat.
The survey for Korea is from the Korean Survey of Lending Attitudes, which
asks about households’ “housing lending.” The diffusion index is the sum
of the responses of significant increase plus responses of moderate increase
minus responses of a significant decrease minus responses of moderate
decrease times 0.5, divided by 100. For Canada, the Balance of Opinion
survey delivers only a net percentage indicator based only on overall lending conditions (inclusive of residential mortgages but also of other forms of
credit). The net percentage indicator we use is minus a weighted percentage
of surveyed financial institutions reporting tightened credit conditions plus
the weighted percentage reporting eased credit conditions.
Finally, when we analyze the international data in panel regressions, for
the United States, we construct a diffusion index (rather than use the net percentage indicator) from the data reported by the Senior Loan Officer Opinion Survey on Bank Lending Practices for the United States, published by
the Federal Reserve. Data for 2007 onwards is a weighted average of prime
and subprime mortgages, with the following weights for prime mortgages:
2007, 0.75; 2008, 0.95; 2009 and 2010, 0.
Data sources for each country are given in table 6A.3.
Estimation Details
This section presents several details pertaining to our empirical estimation.
Estimating Quarterly Net Foreign Holdings of US Assets
This section describes how we estimate quarterly net foreign holdings of
US assets by combining annual positions data. Annual net foreign holdings estimates are compiled by the BEA in their international investment
data, year- end positions, thereby providing annual observations.17 To form
17. The BEA year- end holdings data begin in 1976. This is in contrast to the TIC data on
asset holdings, which is reported annually only starting in 2002. Thus, the BEA constructs its
294
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
Table 6A.3
Austria
Belgium
Canada
Euro Area
France
Ireland
Korea
Netherlands
Portugal
Spain
United States
Data sources for credit standards
Oesterreichische Nationalbank. Bank Lending Survey.
Nationale Bank van Belgie. Bank Lending Survey.
Bank of Canada. Senior Loan Officer Survey. Lending conditions:
Balance of Opinion.
European Central Bank. Bank Lending Survey.
Banque de France. Bank Lending Survey.
Central Bank of Ireland. Bank Lending Survey. Mean.
Bank of Korea, Financial System Review. Survey Bank Lending Practices.
Lending attitude.
De Nederlandsche Bank. Bank Lending Survey.
Banco de Portugal. Bank Lending Survey.
Banco de Espana. Bank Lending Survey.
Federal Reserve. Senior Loan Officer Opinion Survey.
Notes: For countries other than Korea and Canada, surveys follow the BLS survey conducted
by the European Central Bank. In that survey, the questions attained for our purpose are Q8.1
and Q13.1, about mortgage credit. We construct diffusion indices based on this question for
use in the panel regressions. For Korea, we use the “Lending Attitude” diffusion index for
households’ housing, and for Canada we use the “Overall Balance of Opinion” diffusion index.
quarter- end net foreign liability position of the United States in total securities, we employ a methodology to interpolate between the year- end positions, taking into account the quarterly transactions data in these same
securities.18
Let nhQ4 be the value of net foreign holdings of total securities observed
at the end of quarter Q4 of a given year, where net foreign holdings are
defined as foreign holdings of US securities minus US holdings of foreign
securities. These data are available from the BEA year- end positions table.
q be an estimate that we will form of the value of these net foreign
Let nh
holdings at the end of quarter q in that same year. Let ntq be net transactions
in those securities during that quarter, where net transactions are net foreign
purchases (gross foreign purchases less gross foreign sales) of US- owned
securities minus US net purchases (gross US purchases less gross US sales)
of foreign- owned securities. These data are available from the BEA international transactions table. To obtain estimates of quarterly holdings for the
three quarters within a year, we accumulate according to
= nh
+ nt + adj
nh
q
q –1
q
q
where
own estimate of year- end positions prior to 2002 using as raw inputs the TIC flows data and
the periodic TIC benchmark surveys of holdings. The BEA year- end data are located at http://
www.bea.gov/international/xls/intinv10_t2.xls
18. The data on international transactions in financial securities are in the balance of payments data set, found at http://www.bea.gov/international/xls/table1.xls.
International Capital Flows and House Prices
295
adjq = gapQ4 ∗ wq,Q4
wq,Q4 ≡
ntq
∑
3
k =0
ntQ 4− k
Σ
gapQ4 ≡ (nhQ4 − nhQ4−4) −
3
k=0
ntQ4−k.
The above recursion ensures that our estimate of the holdings at the end
Q 4, is equal to the recorded value from the annual
of Q4 of a given year, nh
holdings data, nhQ4. For all other quarters within a year, the above recursion
forms an estimate of holdings at the end of the quarter, which is equal to the
estimated net holdings from last quarter, plus the net transactions in that
quarter, plus an adjustment. The adjustment is equal to the gap between the
change in measured holdings from the year in which the quarter resides and
the previous year and the cumulation of all the quarterly transactions over
the year, times a weight, where the weight is given by that quarter’s value of
net transactions relative to the value over the entire year. Thus, quarters for
which net transactions were higher in absolute value receive a greater weight
in the adjustment. Notice that, in the absence of any valuation adjustments,
the cumulation of all the quarterly transactions over the year would equal
the total change in net foreign holdings or year- end positions. The observed
change in year- end positions takes into account the valuation changes, and
gapQ4 is the difference between the observed change in year- end positions
and the cumulation of the quarterly transactions. Thus, roughly speaking,
the adjustment adjq captures the pure valuation effects that are not reflected
in the cumulation of transactions but are reflected in the total change in
net foreign holdings. The weights wq,Q4 give quarters with a larger value of
transactions more weight in the adjustment.
US Regressions
For the contemporaneous quarterly regressions we report Newey–West
corrected standard errors (and t- statistics) using 4 lags. For the long horizon
quarterly regressions we use lags equal to max{Horizon − 1,4}, to take into
consideration the use of overlapping data.
Panel Regressions
We use a balanced panel from 2002:Q4 to 2010:Q4 for ten countries
plus the Euro Area: Austria, Belgium, Canada, Euro Area, France, Ireland, Korea, the Netherlands, Portugal, Spain, United States. The choice of
sample period is determined by the availability of a balanced panel for data
on credit standards (European Central Bank conducts the Bank Lending
Survey since 2002:Q4), and quarterly house prices (for Italy and Germany
we only have annual data on house prices, for Hungary only semiannual data
on credit conditions, and for Poland we only have data on credit conditions
296
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
since 2003:Q4). We also use a subsample of nine countries where we drop
Canada and the Netherlands (see previous information on credit standards).
The Euro Area consists of seventeen countries: Austria, Belgium, Cyprus,
Estonia, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg,
Malta, the Netherlands, Portugal, Slovakia, Slovenia, and Spain. We construct log changes log(HPt) − log(HPt−1) for contemporaneous changes, and
log(HPt+H ) − log(HPt) for H horizon changes.
For the contemporaneous quarterly regressions, we report the robust
standard errors (and t- statistics) using the Driscoll-Kraay statistic, with
lags = 3 (default). For the quarterly long horizon regressions, we use instead
number of lags equal to max{Horizon − 1,3} to take into account the use
of overlapping data. For more on the Driscoll–Kraay statistic, see Hoechle
(2007) and Driscoll and Kraay (1998).
References
Acharya, V. V., T. Cooley, M. Richardson, and I. Walter. 2011. Regulating Wall
Street: The Dodd–Frank Act and the New Architecture of Global Finance. Wiley
Finance Series. Hoboken, NJ: John Wiley and Sons.
Acharya, V. V., and M. Richardson. 2011. Restoring Financial Stability: How to
Repair a Failed System. Wiley Finance Series. Hoboken, NJ: John Wiley and Sons.
Acharya, V. V., M. Richardson, S. Van Nieuwerburgh, and L. J. White. 2011. Guaranteed to Fail: Freddie, Fannie, and the Debacle of US Mortgage Finance. Princeton,
NJ: Princeton University Press.
Acharya, V. V., P. Schnabl, and G. Suarez. 2012. “Securitization without Risk Transfer.” Forthcoming, Journal of Financial Economics.
Adam, K., A. Marcet, and P. Kuang. 2011. “House Price Booms and the Current
Account.” NBER Working Paper no. 17224. Cambridge, MA: National Bureau
of Economic Research.
Adelino, M., A. Schoar, and F. Severino. 2012. “Credit Supply and House Prices:
Evidence from Mortgage Market Segmentation.” NBER Working Paper no.
17832. Cambridge, MA: National Bureau of Economic Research.
Aizenman, J., and Y. Jinjarak. 2009. “Current Account Patterns and National Real
Estate Markets.” Journal of Urban Economics 66 (2): 75– 89.
Alfaro, L., S. Kalemli-Ozcan, and V. Volosovych. 2011. “Sovereigns, Upstream
Capital Flows, and Global Imbalances.” Unpublished Paper, Harvard Business
School.
Amromin, G., J. Huang, C. Sialm, and E. Zhong. 2011. “Complex Mortgages.”
NBER Working Paper no. 17315. Cambridge, MA: National Bureau of Economic
Research.
Barlevy, G., and J. D. Fisher. 2011. “Mortgage Choices and Housing Speculation.”
Federal Reserve Bank of Chicago, Working Paper no. 2010-12.
Bernanke, B. S. 2005. “Remarks by Governor Ben S. Bernanke at the Sandridge
Lecture.” Virginia Association of Economics. Richmond, Virginia, March 10.
———. 2008. “Remarks by Chairman Ben S. Bernanke at the International Monetary Conference.” Barcelona, Spain (via satellite), June 3.
International Capital Flows and House Prices
297
———. 2011. “International Capital Flows and the Returns to Safe Assets in the
United States 2003– 2007.” Banque de France Financial Stability Review 15:13– 26.
Bernanke, B. S., and M. Gertler. 1989. “Agency Costs, Net Worth and Business Cycle
Flutuations.” American Economic Review 79:14– 31.
Bernanke, B. S., V. R. Reinhart, and B. P. Sack. 2004. “Monetary Policy Alternatives
at the Zero Bound: An Empirical Assessment.” Brookings Papers on Economic
Activity (2):1– 100. Washington, DC: Brookings Institution.
Boz, E., and E. Mendoza. 2010. “Financial Innovation, the Discovery of Risk, and
the US Credit Crisis.” CEPR Discussion Paper no. 7967. Center for Economic
Policy Research.
Caballero, R. J., E. Fahri, and P.-O. Gourinchas. 2008. “An Equilibrium Model of
‘Global Imbalances’ and Low Interest Rates.” American Economic Review 98 (1):
358– 93.
Caballero, R. J., and A. Krishnamurthy. 2009. “Global Imbalances and Financial
Fragility.” American Economic Review Papers and Proceedings 99:584– 88.
Case, K. E., and R. J. Shiller. 1989. “The Efficiency of the Market for Single-Family
Homes.” American Economic Review 79 (1): 125– 37.
Driscoll, J. C., and A. C. Kraay. 1998. “Consistent Covariance Matrix Estimation
with Spatially Dependent Panel Data.” The Review of Economics and Statistics 80
(4): 549– 60.
Favara, G., and J. Imbs. 2011. “Credit Supply and the Price of Housing.” 2011 Meeting Paper no. 1342. Society for Economic Dynamics.
Favilukis, J., S. C. Ludvigson, and S. Van Nieuwerburgh. 2010. “The Macroeconomic
Effects of Housing Wealth, Housing Finance and Limited Risk-Sharing in
General Equilibrium.” NBER Working Paper no. 15988. Cambridge, MA:
National Bureau of Economic Research.
———. 2011. “Foreign Ownership of US Safe Assets: Good or Bad?” Unpublished
Paper, New York University.
Ferrero, A. 2011. “House Prices Booms and Current Account Deficits.” Unpublished
Paper, Federal Reserve Bank of New York.
Geithner, T. 2007. “Remarks by New York Federal Reserve President Timothy
Geithner at the Council on Foreign Relations.” C. Peter McColough Roundtable
Series on International Economics. New York, January 11.
Gete, P. 2010. “Housing Markets and Current Account Dynamics.” Available at
SSRN: http://ssrn.com/abstract=1558512. Social Science Research Network.
Haughwout, A., J. Tracy, R. Peach, and E. Okah. 2011. “Mortgages Defaults and
the Credit Crunch.” Federal Reserve Bank of New York Unpublished Paper.
Himmelberg, C., C. Mayer, and T. Sinai. 2005. “Assessing High House Prices:
Bubbles, Fundamentals and Misperceptions.” Journal of Economic Perspectives
19 (4): 67– 92.
Hodrick, R., and E. C. Prescott. 1997. “Post-War US Business Cycles: A Descriptive
Empirical Investigation.” Journal of Money, Credit, and Banking 29:1– 16.
Hoechle, Daniel. 2007. “Robust Standard Errors for Panel Regressions with CrossSectional Dependence.” Available at http://fmwww.bc.edu/repec/bocode/x/xtscc
_paper.pdf.
Kiyotaki, N., and J. Moore. 1997. “Credit Cycles.” Journal of Political Economy 105
(2): 211– 48.
Kohn, D. L. 2002. “Panel: Implications of Declining Treasury Debt. What Should
the Federal Reserve Do as Treasury Debt Is Repaid?” Journal of Money, Credit,
and Banking 34 (3): 941– 45.
Kole, L. S., and R. F. Martin. 2009. “The Relationship between House Prices and
the Current Account.” Working Paper, Board of Governors of the Federal Reserve
System.
298
J. Favilukis, D. Kohn, S. C. Ludvigson, and S. Van Nieuwerburgh
Krainer, J., S. F. LeRoy, and O. Munpyung. 2009. “Mortgage Default and Mortgage
Valuation.” Working Paper Series 2009– 20. San Francisco: Federal Reserve Bank
of San Francisco.
Krishnamurthy, A., and A. Vissing-Jorgensen. 2007. “The Demand for Treasury
Debt.” NBER Working Paper no. 12881. Cambridge, MA: National Bureau of
Economic Research.
Laibson, D., and J. Mollerstrom. 2010. “Capital Flows, Consumption Booms and
Asset Bubbles: A Behavioural Alternative to the Savings Glut Hypothesis.” Economic Journal 120:354– 74.
Laufer, S. 2011. “Equity Extraction and Mortgage Default.” Working Paper, New
York University.
Maddaloni, A., and J.-L. Peydro. 2011. “Bank Risk-Taking, Securitization, Supervision, and Low Interest Rates: Evidence from the Euro-Area and the US Lending
Standards.” Review of Financial Studies 24 (6): 2121– 65.
Mendoza, E. G., V. Quadrini, and J.-V. Rios-Rull. 2007. “Financial Integration,
Financial Deepness and Global Imbalances.” NBER Working Paper no. 12909.
Cambridge, MA: National Bureau of Economic Research.
Mian, A., and A. Sufi. 2009. “The Consequences of Mortgage Expansion: Evidence
from the US Mortgage Default Crisis.” Quarterly Journal of Economics 124 (4):
1449– 96.
Mian, A., A. Sufi, and F. Trebbi. 2010. “The Political Economy of the US Mortgage
Default Crisis.” American Economic Review 100:67– 98.
Newey, W. K., and K. D. West. 1987. “A Simple, Positive Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix.” Econometrica
55:703– 38.
Nichols, J., A. Pennington-Cross, and A. Yezer. 2005. “Borrower Self-Selection,
Underwriting Costs, and Subprime Mortgage Credit Supply.” Journal of Real
Estate Finance and Economics 30:197– 219.
Obstfeld, M. 2011. “Financial Flows, Financial Crises, and Global Imbalances.”
Keynote address to the 15th International Conference on Macroeconomic Analysis and International Finance. University of Crete, Greece, May 26– 28.
Piskorski, T., and A. Tchistyi. 2010. “Optimal Mortgage Design.” Review of Financial Studies 23:3098– 140.
Poon, M. A. 2009. “From New Deal Institutions to Capital Markets: Commercial
Consumer Risk Scores and the Making of Subprime Mortgage Finance.” Accounting, Organizations and Society 35 (5): 654– 74.
Rajan, R., and R. Ramcharan. 2011a. “The Anatomy of a Credit Crisis: The Boom
and Bust in Farm Land Prices in the United States in the 1920s.” In PIEP: Research
Group on Political Institutions and Economic Policy. Cambridge, MA: Weatherhead Center for International Affairs, Institute for Quantitative Social Sciences.
Available at http://conferences.wcfia.harvard.edu/piep/publications/anatomy
- credit- crisis- boom- and- bust- farm- land- prices- united- states- 1920s.
———. 2011b. “Constituencies and Legislation: The Fight over the McFadden Act
of 1927.” NBER Working Paper no. 17266. Cambridge, MA: National Bureau of
Economic Research.
Rice, T., and P. E. Strahan. 2010. “Does Credit Competition Affect Small-Firm
Finance?” The Journal of Finance 65 (3): 861– 89.
Streitfeld, D. 2009. “Tight Mortgage Rules Exclude Even Good Risks.” New York
Times, Saturday, July 11.
Taylor, J. B. 2009. Getting off Track: How Government Actions and Interventions
Caused, Prolonged, and Worsened the Financial Crisis. Stanford: Hoover Institution Press.
International Capital Flows and House Prices
299
US Treasury Department. 2011. Report on Foreign Portfolio Holdings of US Securities, as of June 30, 2010. Department of Treasury, Federal Reserve Bank of New
York, Board of Governors of the Federal Reserve System.
Warnock, F. 2010. “Two Myths About the U.S. Dollar.” Center for Geoeconomic
Studies Capital Flows Quarterly. Council on Foreign Relations Press, September.
Warnock, F. E., and V. C. Warnock. 2009 “International Capital Flows and US
Interest Rates.” Journal of International Money and Finance 28:903– 19.